Research planning · November 2026
AI Model Projects & Paper Initiatives
One distinct base family or method per proposed project and paper initiative. These linked working records are planning templates, not completed software or published research. Return to the distinct-types register.
Reconstructing institutional decisions from incomplete recordsAI evaluation · Study design in development
Project: Compare direct reconstruction, claim-level citations, and citations with explicit gap marking using fictional case files with known decision sequences, missing records, contradictions and corrections.
Research question: Does requiring source links and explicit gap marking reduce unsupported claims while preserving supported facts?
Paper initiative: Technical evaluation of evidence-grounded institutional decision reconstruction.
Measure unsupported claims, whether citations support their claims, and how many supported facts are retained. Study design in development.
Convolutional neural networkAI evaluation · Study design in development
AIT-0001 · model architecture · Computer vision
Project: Evaluate Convolutional neural network against a documented reference baseline
Test environment: a controlled image-processing dataset with annotation and shift testing.
Research question: Under which evidence conditions does Convolutional neural network improve or fail at local spatial receptive fields?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Convolutional neural network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Vision TransformerAI evaluation · Study design in development
AIT-0002 · model architecture · Computer vision
Project: Evaluate Vision Transformer against a documented reference baseline
Test environment: a controlled image-processing dataset with annotation and shift testing.
Research question: Under which evidence conditions does Vision Transformer improve or fail at patch-token global attention?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Vision Transformer in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: CAT-016 — Object detection task. Research history retained.
AIT-0003 · model architecture class · Computer vision
Historical: Object detection network
Project: Evaluate Object detection network against a documented reference baseline
Test environment: a controlled image-processing dataset with annotation and shift testing.
Research question: Under which evidence conditions does Object detection network improve or fail at object localisation and identity?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Object detection network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: CAT-017 — Semantic segmentation task. Research history retained.
AIT-0004 · model architecture class · Computer vision
Historical: Semantic segmentation network
Project: Evaluate Semantic segmentation network against a documented reference baseline
Test environment: a controlled image-processing dataset with annotation and shift testing.
Research question: Under which evidence conditions does Semantic segmentation network improve or fail at dense pixel-level category assignment?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Semantic segmentation network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: CAT-018 — Instance segmentation task. Research history retained.
AIT-0005 · model architecture class · Computer vision
Historical: Instance segmentation network
Project: Evaluate Instance segmentation network against a documented reference baseline
Test environment: a controlled image-processing dataset with annotation and shift testing.
Research question: Under which evidence conditions does Instance segmentation network improve or fail at separate masks for object instances?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Instance segmentation network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Encoder TransformerAI evaluation · Study design in development
AIT-0006 · model architecture · Language and knowledge
Project: Evaluate Encoder Transformer against a documented reference baseline
Test environment: a source-linked text and reasoning benchmark with human-reviewed answers.
Research question: Under which evidence conditions does Encoder Transformer improve or fail at contextual bidirectional language representation?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Encoder Transformer in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Autoregressive TransformerAI evaluation · Study design in development
AIT-0007 · model architecture · Language and knowledge
Project: Evaluate Autoregressive Transformer against a documented reference baseline
Test environment: a source-linked text and reasoning benchmark with human-reviewed answers.
Research question: Under which evidence conditions does Autoregressive Transformer improve or fail at next-token generation with long-context dependencies?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Autoregressive Transformer in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Retrieval-augmented generationAI evaluation · Study design in development
AIT-0008 · system integration pattern · Language and knowledge
Project: Evaluate Retrieval-augmented generation against a documented reference baseline
Test environment: a source-linked text and reasoning benchmark with human-reviewed answers.
Research question: Under which evidence conditions does Retrieval-augmented generation improve or fail at retrieval-to-answer grounding relationships?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Retrieval-augmented generation in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Sequence-to-sequence TransformerAI evaluation · Study design in development
AIT-0009 · model architecture · Language and knowledge
Project: Evaluate Sequence-to-sequence Transformer against a documented reference baseline
Test environment: a source-linked text and reasoning benchmark with human-reviewed answers.
Research question: Under which evidence conditions does Sequence-to-sequence Transformer improve or fail at input-output sequence transformations?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Sequence-to-sequence Transformer in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Knowledge-graph question answeringAI evaluation · Study design in development
AIT-0010 · application architecture · Language and knowledge
Project: Evaluate Knowledge-graph question answering against a documented reference baseline
Test environment: a source-linked text and reasoning benchmark with human-reviewed answers.
Research question: Under which evidence conditions does Knowledge-graph question answering improve or fail at entity-link and graph-path evidence?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Knowledge-graph question answering in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Gradient-boosted treesAI evaluation · Study design in development
AIT-0011 · machine-learning algorithm family · Predictive and tabular
Project: Evaluate Gradient-boosted trees against a documented reference baseline
Test environment: a synthetic time-series or tabular dataset with holdout and calibration evaluation.
Research question: Under which evidence conditions does Gradient-boosted trees improve or fail at nonlinear feature-outcome interactions?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Gradient-boosted trees in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Generalised additive modelAI evaluation · Study design in development
AIT-0012 · statistical model family · Predictive and tabular
Project: Evaluate Generalised additive model against a documented reference baseline
Test environment: a synthetic time-series or tabular dataset with holdout and calibration evaluation.
Research question: Under which evidence conditions does Generalised additive model improve or fail at additive, interpretable predictors?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Generalised additive model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Temporal forecasting TransformerAI evaluation · Study design in development
AIT-0013 · model architecture · Predictive and tabular
Project: Evaluate Temporal forecasting Transformer against a documented reference baseline
Test environment: a synthetic time-series or tabular dataset with holdout and calibration evaluation.
Research question: Under which evidence conditions does Temporal forecasting Transformer improve or fail at time-varying series dependencies?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Temporal forecasting Transformer in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
State-space sequence modelAI evaluation · Study design in development
AIT-0014 · model architecture · Predictive and tabular
Project: Evaluate State-space sequence model against a documented reference baseline
Test environment: a synthetic time-series or tabular dataset with holdout and calibration evaluation.
Research question: Under which evidence conditions does State-space sequence model improve or fail at latent evolving state relationships?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of State-space sequence model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Cox proportional hazards modelAI evaluation · Study design in development
AIT-0015 · semiparametric statistical model · Time-to-event analysis
Project: Evaluate Cox proportional hazards regression against a baseline survival model on synthetic time-to-event data with documented censoring and proportional-hazards assumptions.
Research question: Under which conditions does the Cox model yield stable and calibrated risk estimates, and how do assumption violations affect its interpretation?
Project status: Proposed; no implementation or benchmark completed. Distinct from the broader survival-analysis methodology ALG-009.
Paper initiative: The Cox proportional hazards model: assumptions, limitations and evaluation against survival-analysis alternatives.
Status: proposed manuscript; experiments and systematic literature review outstanding.
Cox (1972), foundational research · Canonical model record
Graph convolutional networkAI evaluation · Study design in development
AIT-0016 · model architecture · Graphs and relational data
Project: Evaluate Graph convolutional network against a documented reference baseline
Test environment: a synthetic linked-entity graph with labelled relationship tests.
Research question: Under which evidence conditions does Graph convolutional network improve or fail at neighbourhood topology aggregation?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Graph convolutional network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Graph attention networkAI evaluation · Study design in development
AIT-0017 · model architecture · Graphs and relational data
Project: Evaluate Graph attention network against a documented reference baseline
Test environment: a synthetic linked-entity graph with labelled relationship tests.
Research question: Under which evidence conditions does Graph attention network improve or fail at attention-weighted neighbourhood relationships?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Graph attention network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Relational graph neural networkAI evaluation · Study design in development
AIT-0018 · model architecture · Graphs and relational data
Project: Evaluate Relational graph neural network against a documented reference baseline
Test environment: a synthetic linked-entity graph with labelled relationship tests.
Research question: Under which evidence conditions does Relational graph neural network improve or fail at typed edges and heterogeneous relations?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Relational graph neural network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Temporal graph neural networkAI evaluation · Study design in development
AIT-0019 · model architecture · Graphs and relational data
Project: Evaluate Temporal graph neural network against a documented reference baseline
Test environment: a synthetic linked-entity graph with labelled relationship tests.
Research question: Under which evidence conditions does Temporal graph neural network improve or fail at time-stamped dynamic graph changes?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Temporal graph neural network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. No canonical replacement assigned; definition requires evidence review. Research history retained.
AIT-0020 · research model/system concept · Graphs and relational data
Historical: Neural relational database model
Project: Evaluate Neural relational database model against a documented reference baseline
Test environment: a synthetic linked-entity graph with labelled relationship tests.
Research question: Under which evidence conditions does Neural relational database model improve or fail at foreign-key and multi-table entity relations?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Neural relational database model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Denoising diffusion modelAI evaluation · Study design in development
AIT-0021 · generative model architecture · Generative and multimodal
Project: Evaluate Denoising diffusion model against a documented reference baseline
Test environment: a consented synthetic multimodal benchmark with output-quality and safety checks.
Research question: Under which evidence conditions does Denoising diffusion model improve or fail at noise-to-sample denoising trajectories?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Denoising diffusion model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Latent diffusion modelAI evaluation · Study design in development
AIT-0022 · generative model architecture · Generative and multimodal
Project: Evaluate Latent diffusion model against a documented reference baseline
Test environment: a consented synthetic multimodal benchmark with output-quality and safety checks.
Research question: Under which evidence conditions does Latent diffusion model improve or fail at compressed latent-to-output relationships?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Latent diffusion model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Generative adversarial networkAI evaluation · Study design in development
AIT-0023 · generative model architecture · Generative and multimodal
Project: Evaluate Generative adversarial network against a documented reference baseline
Test environment: a consented synthetic multimodal benchmark with output-quality and safety checks.
Research question: Under which evidence conditions does Generative adversarial network improve or fail at generator-discriminator training dynamics?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Generative adversarial network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Variational autoencoderAI evaluation · Study design in development
AIT-0024 · generative model architecture · Generative and multimodal
Project: Evaluate Variational autoencoder against a documented reference baseline
Test environment: a consented synthetic multimodal benchmark with output-quality and safety checks.
Research question: Under which evidence conditions does Variational autoencoder improve or fail at latent distribution and reconstruction relations?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Variational autoencoder in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Vision-language modelAI evaluation · Study design in development
AIT-0025 · multimodal model architecture class · Generative and multimodal
Project: Evaluate Vision-language model against a documented reference baseline
Test environment: a consented synthetic multimodal benchmark with output-quality and safety checks.
Research question: Under which evidence conditions does Vision-language model improve or fail at cross-modal image-text correspondences?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Vision-language model in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Tool-using LLM agentAI evaluation · Study design in development
AIT-0026 · agent system architecture · Agents and automation
Project: Evaluate Tool-using LLM agent against a documented reference baseline
Test environment: a sandboxed workflow with a permissions ledger and human approval.
Research question: Under which evidence conditions does Tool-using LLM agent improve or fail at model-to-tool requests and permissions?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Tool-using LLM agent in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Plan-and-execute agentAI evaluation · Study design in development
AIT-0027 · agent orchestration pattern · Agents and automation
Project: Evaluate Plan-and-execute agent against a documented reference baseline
Test environment: a sandboxed workflow with a permissions ledger and human approval.
Research question: Under which evidence conditions does Plan-and-execute agent improve or fail at goal decomposition and action dependencies?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Plan-and-execute agent in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Multi-agent orchestrationAI evaluation · Study design in development
AIT-0028 · multi-agent system pattern · Agents and automation
Project: Evaluate Multi-agent orchestration against a documented reference baseline
Test environment: a sandboxed workflow with a permissions ledger and human approval.
Research question: Under which evidence conditions does Multi-agent orchestration improve or fail at delegation and message provenance?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Multi-agent orchestration in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: CAT-015 — AI-assisted process automation. Research history retained.
AIT-0029 · business workflow architecture · Agents and automation
Historical: AI-enabled BPM workflow
Project: Evaluate AI-enabled BPM workflow against a documented reference baseline
Test environment: a sandboxed workflow with a permissions ledger and human approval.
Research question: Under which evidence conditions does AI-enabled BPM workflow improve or fail at tasks, transitions and business rules?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of AI-enabled BPM workflow in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Event-driven decision agentAI evaluation · Study design in development
AIT-0030 · agent system architecture · Agents and automation
Project: Evaluate Event-driven decision agent against a documented reference baseline
Test environment: a sandboxed workflow with a permissions ledger and human approval.
Research question: Under which evidence conditions does Event-driven decision agent improve or fail at event conditions and triggered actions?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Event-driven decision agent in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: CAT-008 — Robotics and embodied AI. Research history retained.
AIT-0031 · robotics policy architecture · Robotics and control
Historical: Vision-based robotic policy
Project: Evaluate Vision-based robotic policy against a documented reference baseline
Test environment: a simulation-only control task with reproducible trajectories and safety constraints.
Research question: Under which evidence conditions does Vision-based robotic policy improve or fail at visual input to motor action?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Vision-based robotic policy in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Model-predictive controllerAI evaluation · Study design in development
AIT-0032 · control method · Robotics and control
Project: Evaluate Model-predictive controller against a documented reference baseline
Test environment: a simulation-only control task with reproducible trajectories and safety constraints.
Research question: Under which evidence conditions does Model-predictive controller improve or fail at future-state planning under constraints?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Model-predictive controller in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: PAR-003 — Reinforcement learning. Research history retained.
AIT-0033 · learning-policy method · Robotics and control
Historical: Reinforcement learning policy
Project: Evaluate Reinforcement learning policy against a documented reference baseline
Test environment: a simulation-only control task with reproducible trajectories and safety constraints.
Research question: Under which evidence conditions does Reinforcement learning policy improve or fail at state-action-reward dynamics?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Reinforcement learning policy in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Archived / Superseded · Historical project formulation; not an active distinct AI architecture. Canonical replacement: PAR-005 — Imitation learning. Research history retained.
AIT-0034 · learning-policy method · Robotics and control
Historical: Imitation learning policy
Project: Evaluate Imitation learning policy against a documented reference baseline
Test environment: a simulation-only control task with reproducible trajectories and safety constraints.
Research question: Under which evidence conditions does Imitation learning policy improve or fail at demonstration-to-action generalisation?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Imitation learning policy in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Spiking neural controllerAI evaluation · Study design in development
AIT-0035 · neural control architecture · Robotics and control
Project: Evaluate Spiking neural controller against a documented reference baseline
Test environment: a simulation-only control task with reproducible trajectories and safety constraints.
Research question: Under which evidence conditions does Spiking neural controller improve or fail at event-driven neural timing dynamics?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Spiking neural controller in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Rule-based expert systemAI evaluation · Study design in development
AIT-0036 · symbolic AI system · Logic, search and hybrid decision
Project: Evaluate Rule-based expert system against a documented reference baseline
Test environment: a documented rule-and-inference problem set with counterexamples.
Research question: Under which evidence conditions does Rule-based expert system improve or fail at if-then inferences and exception rules?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Rule-based expert system in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Constraint-satisfaction solverAI evaluation · Study design in development
AIT-0037 · constraint-solving algorithm · Logic, search and hybrid decision
Project: Evaluate Constraint-satisfaction solver against a documented reference baseline
Test environment: a documented rule-and-inference problem set with counterexamples.
Research question: Under which evidence conditions does Constraint-satisfaction solver improve or fail at variables, constraints and feasible assignments?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Constraint-satisfaction solver in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Bayesian networkAI evaluation · Study design in development
AIT-0038 · probabilistic graphical model · Logic, search and hybrid decision
Project: Evaluate Bayesian network against a documented reference baseline
Test environment: a documented rule-and-inference problem set with counterexamples.
Research question: Under which evidence conditions does Bayesian network improve or fail at conditional probabilistic dependence?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Bayesian network in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Neuro-symbolic reasoning systemAI evaluation · Study design in development
AIT-0039 · hybrid reasoning architecture · Logic, search and hybrid decision
Project: Evaluate Neuro-symbolic reasoning system against a documented reference baseline
Test environment: a documented rule-and-inference problem set with counterexamples.
Research question: Under which evidence conditions does Neuro-symbolic reasoning system improve or fail at learned representations with explicit symbolic rules?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Neuro-symbolic reasoning system in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Case-based reasoning systemAI evaluation · Study design in development
AIT-0040 · reasoning method · Logic, search and hybrid decision
Project: Evaluate Case-based reasoning system against a documented reference baseline
Test environment: a documented rule-and-inference problem set with counterexamples.
Research question: Under which evidence conditions does Case-based reasoning system improve or fail at analogy and difference across past cases?
Project status: Proposed; implementation and evaluation outstanding.
Paper initiative: Evidence and failure modes of Case-based reasoning system in a controlled comparison
Status: proposed manuscript / literature and experimental evidence not yet assembled. Distinguish the architecture or method from the training and evaluation regimes used with it.
Wikipedia context · Type register
Statistical and learning-method initiatives
These are methodological projects, not separately established neural architecture types.
Survival analysisAI evaluation · Study design in development
M-001 · Statistical time-to-event methodology · Proposed method project
Project: Investigate time-to-event modelling with censored synthetic observations
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Compare Kaplan–Meier, Cox PH and ML-based survival methods using held-out censoring tests.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Fuzzy logicAI evaluation · Study design in development
M-002 · Logic and approximate reasoning · Proposed method project
Project: Build a synthetic fuzzy-rule inference example
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Compare crisp-rule and fuzzy-rule reasoning on documented uncertainty cases.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Supervised learningAI evaluation · Study design in development
M-003 · Machine-learning training paradigm · Proposed method project
Project: Evaluate supervised learning on labelled synthetic data
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Compare generalisation and calibration under shifted labels.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Unsupervised learningAI evaluation · Study design in development
M-004 · Machine-learning training paradigm · Proposed method project
Project: Evaluate clustering and representation methods on unlabelled synthetic data
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Compare stability under preprocessing changes and label-free validity checks.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Reinforcement learningAI evaluation · Study design in development
M-005 · Sequential learning paradigm · Proposed method project
Project: Test agent learning in a controlled simulated environment
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Compare learned policies against scripted baselines with reproducible reward scenarios.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Deep learningAI evaluation · Study design in development
M-006 · Machine-learning approach · Proposed method project
Project: Compare a multilayer learner against simpler baselines
Test status: Methodological experiment proposed; not implemented or validated.
Paper initiative: Examine representation learning, data requirements and failure modes.
Manuscript proposed; no publication or performance claims.
Methodological reference · Method table
Classification review · new project and paper initiatives
Proposed records only. Canonical types are distinguished from historical labels.
Object detectionAI evaluation · Study design in development
CAT-016 · Computer vision task · Proposed
Project initiative: Compare Object detection with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Object detection.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Semantic segmentationAI evaluation · Study design in development
CAT-017 · Computer vision task · Proposed
Project initiative: Compare Semantic segmentation with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Semantic segmentation.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Instance segmentationAI evaluation · Study design in development
CAT-018 · Computer vision task · Proposed
Project initiative: Compare Instance segmentation with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Instance segmentation.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Imitation learningAI evaluation · Study design in development
PAR-005 · Learning from demonstration paradigm · Proposed
Project initiative: Compare Imitation learning with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Imitation learning.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Artificial neural networkAI evaluation · Study design in development
ARC-033 · Umbrella model architecture family · Proposed
Project initiative: Compare Artificial neural network with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Artificial neural network.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Faster R-CNNAI evaluation · Study design in development
ARC-034 · Region proposal object-detection model architecture · Proposed
Project initiative: Compare Faster R-CNN with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Faster R-CNN.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Fully convolutional network (FCN)AI evaluation · Study design in development
ARC-035 · Dense-prediction model architecture · Proposed
Project initiative: Compare Fully convolutional network (FCN) with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Fully convolutional network (FCN).
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Mask R-CNNAI evaluation · Study design in development
ARC-036 · Instance-segmentation model architecture · Proposed
Project initiative: Compare Mask R-CNN with at least one independently documented baseline. Confirm datasets, evaluation criteria and constraints before implementation.
Research status: Proposed; neither implemented nor validated.
Paper initiative: Taxonomy, evaluation and reproducibility of Mask R-CNN.
Working topic only; literature review, falsification, results and publication are not complete.
Taxonomy background · Canonical register
Self-supervised learningAI evaluation · Study design in development
PAR-004 · Learning paradigm · Proposed
Project initiative: Compare self-supervised representation learning against a supervised baseline using a documented dataset, transparent splits and independent evaluation.
Project status: Proposed; no training or benchmark completed.
Paper initiative: Evaluation of self-supervised learning under documented data and transfer conditions.
Proposed manuscript; literature and experimental evidence outstanding.
Introductory context · Canonical taxonomy entry
Documented architecture and algorithm initiatives · 2026 research expansion
These projects and paper initiatives are proposals linked to original external research. They are not existing software deployments or completed publications.
Residual network (ResNet)AI evaluation · Study design in development
ARC-037 · Residual convolutional network architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Residual network (ResNet) in the domain of Computer vision; residual/skip-connected deep network.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Residual network (ResNet).
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2015) · Canonical record
U-NetAI evaluation · Study design in development
ARC-038 · Encoder–decoder segmentation architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for U-Net in the domain of Dense image segmentation; contracting/expanding paths and skip connections.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of U-Net.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2015) · Canonical record
Long short-term memory networkAI evaluation · Study design in development
ARC-039 · Gated recurrent neural-network architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Long short-term memory network in the domain of Long-range sequential dependencies.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Long short-term memory network.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (1997) · Canonical record
BERTAI evaluation · Study design in development
ARC-040 · Bidirectional Transformer encoder model · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for BERT in the domain of Bidirectional pretrained language representations.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of BERT.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2018) · Canonical record
Detection Transformer (DETR)AI evaluation · Study design in development
ARC-041 · Transformer-based end-to-end object detection architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Detection Transformer (DETR) in the domain of Set prediction and object queries in object detection.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Detection Transformer (DETR).
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2020) · Canonical record
Mamba selective state-space modelAI evaluation · Study design in development
ARC-042 · Selective state-space neural sequence architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Mamba selective state-space model in the domain of Input-dependent selective sequence modelling.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Mamba selective state-space model.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2023) · Canonical record
Neural Turing MachineAI evaluation · Study design in development
ARC-043 · Differentiable neural external-memory architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Neural Turing Machine in the domain of Memory-augmented sequence and algorithmic tasks.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Neural Turing Machine.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2014) · Canonical record
Neural ordinary differential equationAI evaluation · Study design in development
ARC-044 · Continuous-depth neural model architecture · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Neural ordinary differential equation in the domain of Continuous-time latent dynamics and differentiable ODE solvers.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Neural ordinary differential equation.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2018) · Canonical record
CLIP contrastive vision-language modelAI evaluation · Study design in development
ARC-045 · Contrastive dual-encoder multimodal model · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for CLIP contrastive vision-language model in the domain of Image-text representation learning and zero-shot visual classification.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of CLIP contrastive vision-language model.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2021) · Canonical record
T5 text-to-text TransformerAI evaluation · Study design in development
ARC-046 · Unified text-to-text encoder–decoder model · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for T5 text-to-text Transformer in the domain of Transfer learning for text-to-text tasks.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of T5 text-to-text Transformer.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2019) · Canonical record
Sparsely-gated mixture-of-experts layerAI evaluation · Study design in development
ARC-047 · Conditional-computation architectural component · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Sparsely-gated mixture-of-experts layer in the domain of Sparse expert routing in large neural networks.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Sparsely-gated mixture-of-experts layer.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2017) · Canonical record
Proximal policy optimizationAI evaluation · Study design in development
ALG-012 · On-policy reinforcement-learning optimization algorithm · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Proximal policy optimization in the domain of Clipped policy objective for sequential decision-making.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Proximal policy optimization.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2017) · Canonical record
Soft actor-criticAI evaluation · Study design in development
ALG-013 · Off-policy maximum-entropy actor-critic algorithm · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Soft actor-critic in the domain of Stochastic policy learning and entropy regularization.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Soft actor-critic.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2018) · Canonical record
Deep Q-network learningAI evaluation · Study design in development
ALG-014 · Deep value-based reinforcement-learning algorithm · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Deep Q-network learning in the domain of Q-learning with neural value-function approximator.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Deep Q-network learning.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2013) · Canonical record
XGBoostAI evaluation · Study design in development
ALG-015 · Regularized gradient-tree-boosting implementation and algorithm · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for XGBoost in the domain of Scalable gradient-boosted decision trees.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of XGBoost.
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2016) · Canonical record
Low-rank adaptation (LoRA)AI evaluation · Study design in development
ALG-016 · Parameter-efficient model adaptation method · Working proposal
Project initiative: Design a reproducible implementation and comparison against a documented baseline for Low-rank adaptation (LoRA) in the domain of Low-rank trainable updates to pretrained neural weights.
Research question: Under which conditions does this method or architecture improve performance, reliability, calibration or efficiency compared with simpler alternatives?
Status: Proposed; no implementation, benchmark or result claimed.
Paper initiative: Replication, limitations and comparative evaluation of Low-rank adaptation (LoRA).
Proposed manuscript; literature verification, experimental protocol and independent results outstanding.
Foundational research (2021) · Canonical record
AI taxonomy research expansion · v1.4
Seventeen additional proposed projects and paper initiatives, each cross-referenced to documented external literature. Proposals do not constitute implementation, experiment, manuscript or publication.
Semi-supervised learningAI evaluation · Study design in development
PAR-006 · Learning paradigm combining labelled and unlabelled examples · Proposed
Project: Design a reproducible test or illustrative prototype of Semi-supervised learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Semi-supervised learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2005) · Canonical taxonomy entry
Active learningAI evaluation · Study design in development
PAR-007 · Training paradigm with learner-directed query selection · Proposed
Project: Design a reproducible test or illustrative prototype of Active learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Active learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2009) · Canonical taxonomy entry
Meta-learningAI evaluation · Study design in development
PAR-008 · Learning-to-learn or adaptation-across-tasks paradigm · Proposed
Project: Design a reproducible test or illustrative prototype of Meta-learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Meta-learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2017) · Canonical taxonomy entry
Transfer learningAI evaluation · Study design in development
PAR-009 · Knowledge transfer between learning tasks or domains · Proposed
Project: Design a reproducible test or illustrative prototype of Transfer learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Transfer learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2010) · Canonical taxonomy entry
Federated learningAI evaluation · Study design in development
PAR-010 · Distributed model training with data held across participants · Proposed
Project: Design a reproducible test or illustrative prototype of Federated learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Federated learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2017) · Canonical taxonomy entry
Kaplan–Meier estimatorAI evaluation · Study design in development
ALG-017 · Nonparametric product-limit survival-function estimator · Proposed
Project: Design a reproducible test or illustrative prototype of Kaplan–Meier estimator and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Kaplan–Meier estimator.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (1958) · Canonical taxonomy entry
Random forestAI evaluation · Study design in development
ALG-018 · Bagged decision-tree ensemble learning algorithm · Proposed
Project: Design a reproducible test or illustrative prototype of Random forest and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Random forest.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2001) · Canonical taxonomy entry
Random survival forestAI evaluation · Study design in development
ALG-019 · Survival-tree ensemble for censored time-to-event observations · Proposed
Project: Design a reproducible test or illustrative prototype of Random survival forest and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Random survival forest.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2008) · Canonical taxonomy entry
Support vector machineAI evaluation · Study design in development
ALG-020 · Maximum-margin kernel-based classification method · Proposed
Project: Design a reproducible test or illustrative prototype of Support vector machine and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Support vector machine.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (1995) · Canonical taxonomy entry
Q-learningAI evaluation · Study design in development
ALG-021 · Off-policy temporal-difference action-value learning algorithm · Proposed
Project: Design a reproducible test or illustrative prototype of Q-learning and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Q-learning.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (1992) · Canonical taxonomy entry
Adam optimizerAI evaluation · Study design in development
ALG-022 · Adaptive moment-based stochastic gradient optimisation · Proposed
Project: Design a reproducible test or illustrative prototype of Adam optimizer and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Adam optimizer.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2014) · Canonical taxonomy entry
Multilayer perceptronAI evaluation · Study design in development
ARC-048 · Layered fully-connected feedforward neural model · Proposed
Project: Design a reproducible test or illustrative prototype of Multilayer perceptron and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Multilayer perceptron.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (1986) · Canonical taxonomy entry
Gated recurrent unitAI evaluation · Study design in development
ARC-049 · Gated recurrent neural architecture · Proposed
Project: Design a reproducible test or illustrative prototype of Gated recurrent unit and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Gated recurrent unit.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2014) · Canonical taxonomy entry
YOLO v1 object detectorAI evaluation · Study design in development
ARC-050 · Single-stage convolutional object-detection architecture · Proposed
Project: Design a reproducible test or illustrative prototype of YOLO v1 object detector and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of YOLO v1 object detector.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2016) · Canonical taxonomy entry
GPT-1 generative pre-trained TransformerAI evaluation · Study design in development
ARC-051 · Decoder-only autoregressive pretrained language model · Proposed
Project: Design a reproducible test or illustrative prototype of GPT-1 generative pre-trained Transformer and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of GPT-1 generative pre-trained Transformer.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2018) · Canonical taxonomy entry
Whisper speech recognition modelAI evaluation · Study design in development
ARC-052 · Multitask encoder-decoder speech Transformer model · Proposed
Project: Design a reproducible test or illustrative prototype of Whisper speech recognition model and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of Whisper speech recognition model.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2022) · Canonical taxonomy entry
AlphaFold 2 structure-prediction systemAI evaluation · Study design in development
ARC-053 · Protein structure inference system with Evoformer and structure module · Proposed
Project: Design a reproducible test or illustrative prototype of AlphaFold 2 structure-prediction system and compare it with an appropriate established baseline. Record assumptions, data provenance, failure conditions and resource costs.
Research status: Proposal only; not implemented or experimentally validated.
Paper initiative: Literature-grounded comparison, replicability and limits of AlphaFold 2 structure-prediction system.
Proposed manuscript only, with no completed findings or publication claims.
Foundational or representative reference (2021) · Canonical taxonomy entry
Classical machine learning and modern architecture expansion · v1.5
Thirteen distinct documented techniques or model families. Project and paper initiatives are proposed only.
Logistic regressionAI evaluation · Study design in development
ALG-023 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Logistic regression in the context of Binary class probability modelling via log odds. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Logistic regression.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1958) · Canonical entry
Linear regressionAI evaluation · Study design in development
ALG-024 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Linear regression in the context of Continuous-outcome prediction through linear covariate relationships. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Linear regression.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2013) · Canonical entry
K-means clusteringAI evaluation · Study design in development
ALG-025 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine K-means clustering in the context of Centroid-based data partitioning into k groups. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of K-means clustering.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1982) · Canonical entry
Principal component analysisAI evaluation · Study design in development
ALG-026 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Principal component analysis in the context of Orthogonal dimensionality reduction maximising retained variance. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Principal component analysis.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1986) · Canonical entry
DBSCANAI evaluation · Study design in development
ALG-027 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine DBSCAN in the context of Density-based clustering with explicit noise assignment. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of DBSCAN.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1996) · Canonical entry
Naive Bayes classifierAI evaluation · Study design in development
ALG-028 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Naive Bayes classifier in the context of Classification with conditional-feature-independence approximation. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Naive Bayes classifier.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1997) · Canonical entry
Upper confidence bounds applied to treesAI evaluation · Study design in development
ALG-029 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Upper confidence bounds applied to trees in the context of Monte Carlo search with upper-confidence tree selection. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Upper confidence bounds applied to trees.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2006) · Canonical entry
K-nearest neighborsAI evaluation · Study design in development
ALG-030 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine K-nearest neighbors in the context of Instance-based prediction using local metric neighbors. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of K-nearest neighbors.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1967) · Canonical entry
AdaBoostAI evaluation · Study design in development
ALG-031 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine AdaBoost in the context of Adaptive ensemble boosting from weighted weak learners. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of AdaBoost.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (1997) · Canonical entry
Gaussian process regressionAI evaluation · Study design in development
ALG-032 · Distinct statistical or machine-learning method · Proposed
Project initiative: Examine Gaussian process regression in the context of Bayesian kernel-based function modelling with predictive uncertainty. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Gaussian process regression.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2006) · Canonical entry
RoBERTaAI evaluation · Study design in development
ARC-054 · Distinct model, architecture or training-configuration family · Proposed
Project initiative: Examine RoBERTa in the context of BERT-style language encoder with a revised pretraining recipe. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of RoBERTa.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2019) · Canonical entry
Diffusion TransformerAI evaluation · Study design in development
ARC-055 · Distinct model, architecture or training-configuration family · Proposed
Project initiative: Examine Diffusion Transformer in the context of Diffusion denoising with a Transformer backbone over image latents. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Diffusion Transformer.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2022) · Canonical entry
Swin TransformerAI evaluation · Study design in development
ARC-056 · Distinct model, architecture or training-configuration family · Proposed
Project initiative: Examine Swin Transformer in the context of Hierarchical shifted-window attention for multiscale vision. Establish reproducible baselines, limitations, falsification criteria, data provenance and independent evaluation before interpreting outcomes.
Project status: Proposed; no implemented project or benchmark results claimed.
Paper initiative: Comparative methods and research limits of Swin Transformer.
Working paper proposal; no publication, experimental or critical full-text audit completed.
Original or representative research (2021) · Canonical entry
Symbolic, search, optimisation and graph AI research · v1.6
Nine additional documented AI methods and model architectures. All projects and paper initiatives are proposed, not completed.
A* searchAI evaluation · Study design in development
ALG-033 · Heuristic best-first path-search algorithm · Proposed
Project: Reconstruct a permitted, reproducible example of A* search for Searches graph paths using accumulated cost and estimated remaining cost. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for A* search.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1968) · Canonical entry
Dijkstra's shortest-path algorithmAI evaluation · Study design in development
ALG-034 · Shortest-path graph search algorithm · Proposed
Project: Reconstruct a permitted, reproducible example of Dijkstra's shortest-path algorithm for Computes minimum-cost routes with nonnegative edge weights. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Dijkstra's shortest-path algorithm.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1959) · Canonical entry
Differential evolutionAI evaluation · Study design in development
ALG-035 · Population-based evolutionary optimisation algorithm · Proposed
Project: Reconstruct a permitted, reproducible example of Differential evolution for Optimises continuous objectives using mutation and recombination of candidate solutions. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Differential evolution.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1997) · Canonical entry
STRIPS planningAI evaluation · Study design in development
ALG-036 · Symbolic goal-directed planning method · Proposed
Project: Reconstruct a permitted, reproducible example of STRIPS planning for Finds action sequences over symbolic world states, preconditions and effects. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for STRIPS planning.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1971) · Canonical entry
Viterbi decodingAI evaluation · Study design in development
ALG-037 · Dynamic-programming maximum-likelihood sequence decoding algorithm · Proposed
Project: Reconstruct a permitted, reproducible example of Viterbi decoding for Infers a most probable discrete state sequence under a Markov model. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Viterbi decoding.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1967) · Canonical entry
Real NVP normalizing flowAI evaluation · Study design in development
ARC-057 · Invertible flow-based density-estimation architecture · Proposed
Project: Reconstruct a permitted, reproducible example of Real NVP normalizing flow for Learns exact-likelihood generative density models using coupling layers. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Real NVP normalizing flow.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (2016) · Canonical entry
Graph Isomorphism NetworkAI evaluation · Study design in development
ARC-058 · Weisfeiler–Lehman-expressive graph neural architecture · Proposed
Project: Reconstruct a permitted, reproducible example of Graph Isomorphism Network for Graph classification through expressive multiset neighborhood aggregation. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Graph Isomorphism Network.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (2018) · Canonical entry
Message Passing Neural NetworkAI evaluation · Study design in development
ARC-059 · Graph-message-passing neural architecture framework · Proposed
Project: Reconstruct a permitted, reproducible example of Message Passing Neural Network for Molecular and graph representations via edge messages and node updates. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Message Passing Neural Network.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (2017) · Canonical entry
Baum–Welch algorithmAI evaluation · Study design in development
ALG-038 · Expectation-maximisation training algorithm for hidden Markov models · Proposed
Project: Reconstruct a permitted, reproducible example of Baum–Welch algorithm for Estimates hidden Markov model parameters from incomplete observed sequences. Specify baseline, falsification tests, evaluation data, reproducibility and limitations.
Status: Proposal only; implementation and results outstanding.
Paper initiative: Documented foundations, comparative evaluation and failure conditions for Baum–Welch algorithm.
Unpublished research proposal; no manuscript or completed findings asserted.
Original literature (1970) · Canonical entry
Automated planning and schedulingAI evaluation · Study design in development
CAT-019 · Planning application domain · Proposed
Project: Compare Automated planning and scheduling against appropriate established baselines for goal-directed action sequencing and resource-constrained scheduling; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Automated planning and scheduling.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1971) · Canonical entry
Causal discovery and causal decision supportAI evaluation · Study design in development
CAT-020 · Causal inference application domain · Proposed
Project: Compare Causal discovery and causal decision support against appropriate established baselines for recovering and evaluating causal relationships under explicit assumptions; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Causal discovery and causal decision support.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2000) · Canonical entry
Online learningAI evaluation · Study design in development
PAR-011 · Sequential learning paradigm · Proposed
Project: Compare Online learning against appropriate established baselines for updating predictors incrementally as examples arrive; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Online learning.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1998) · Canonical entry
Continual learningAI evaluation · Study design in development
PAR-012 · Sequential adaptation paradigm · Proposed
Project: Compare Continual learning against appropriate established baselines for learning successive tasks while evaluating retention and forgetting; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Continual learning.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2016) · Canonical entry
Multi-task learningAI evaluation · Study design in development
PAR-013 · Joint-task learning paradigm · Proposed
Project: Compare Multi-task learning against appropriate established baselines for learning related tasks jointly with shared representations; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Multi-task learning.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1997) · Canonical entry
Breadth-first searchAI evaluation · Study design in development
ALG-039 · Uninformed graph search algorithm · Proposed
Project: Compare Breadth-first search against appropriate established baselines for exploring nodes in order of increasing path depth; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Breadth-first search.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1959) · Canonical entry
Depth-first searchAI evaluation · Study design in development
ALG-040 · Uninformed graph search algorithm · Proposed
Project: Compare Depth-first search against appropriate established baselines for exploring paths to terminal depth before backtracking; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Depth-first search.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1972) · Canonical entry
Minimax searchAI evaluation · Study design in development
ALG-041 · Adversarial game-tree algorithm · Proposed
Project: Compare Minimax search against appropriate established baselines for selecting moves by optimising against an opposing decision maker; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Minimax search.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1928) · Canonical entry
Alpha-beta pruningAI evaluation · Study design in development
ALG-042 · Adversarial search acceleration · Proposed
Project: Compare Alpha-beta pruning against appropriate established baselines for pruning game-tree branches that cannot change minimax outcome; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Alpha-beta pruning.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1975) · Canonical entry
Monte Carlo tree searchAI evaluation · Study design in development
ALG-043 · Simulation-guided tree search · Proposed
Project: Compare Monte Carlo tree search against appropriate established baselines for balancing simulated rollouts and exploration of action branches; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Monte Carlo tree search.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2012) · Canonical entry
Simulated annealingAI evaluation · Study design in development
ALG-044 · Stochastic optimisation method · Proposed
Project: Compare Simulated annealing against appropriate established baselines for using probabilistic uphill moves with cooling schedules; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Simulated annealing.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1983) · Canonical entry
Genetic algorithmAI evaluation · Study design in development
ALG-045 · Population-based evolutionary optimisation · Proposed
Project: Compare Genetic algorithm against appropriate established baselines for selection crossover and mutation across candidate solutions; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Genetic algorithm.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1975) · Canonical entry
Particle swarm optimisationAI evaluation · Study design in development
ALG-046 · Swarm-based optimisation · Proposed
Project: Compare Particle swarm optimisation against appropriate established baselines for iterative candidate movement guided by individual and collective best states; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Particle swarm optimisation.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1995) · Canonical entry
Ant colony optimisationAI evaluation · Study design in development
ALG-047 · Swarm-based combinatorial optimisation · Proposed
Project: Compare Ant colony optimisation against appropriate established baselines for pheromone-inspired search over constructive solution paths; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Ant colony optimisation.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1997) · Canonical entry
Expectation-maximisation algorithmAI evaluation · Study design in development
ALG-048 · Latent-variable statistical estimation · Proposed
Project: Compare Expectation-maximisation algorithm against appropriate established baselines for alternating expected latent sufficient statistics and parameter maximisation; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Expectation-maximisation algorithm.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1977) · Canonical entry
Belief propagationAI evaluation · Study design in development
ALG-049 · Probabilistic graphical-model inference · Proposed
Project: Compare Belief propagation against appropriate established baselines for passing local messages to compute marginal or approximate beliefs; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Belief propagation.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2000) · Canonical entry
Markov chain Monte CarloAI evaluation · Study design in development
ALG-050 · Sampling-based Bayesian inference family · Proposed
Project: Compare Markov chain Monte Carlo against appropriate established baselines for constructing markov chains targeting probability distributions; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Markov chain Monte Carlo.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1953) · Canonical entry
Gibbs samplingAI evaluation · Study design in development
ALG-051 · Conditional-distribution MCMC algorithm · Proposed
Project: Compare Gibbs sampling against appropriate established baselines for drawing each variable from its conditional distribution; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Gibbs sampling.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1990) · Canonical entry
Variational inferenceAI evaluation · Study design in development
ALG-052 · Optimisation-based approximate inference · Proposed
Project: Compare Variational inference against appropriate established baselines for approximating intractable distributions using tractable variational families; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Variational inference.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2017) · Canonical entry
Do-calculusAI evaluation · Study design in development
ALG-053 · Causal identification calculus · Proposed
Project: Compare Do-calculus against appropriate established baselines for transforming interventional and observational expressions under causal graphs; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Do-calculus.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1995) · Canonical entry
Propensity score matchingAI evaluation · Study design in development
ALG-054 · Observational causal estimation method · Proposed
Project: Compare Propensity score matching against appropriate established baselines for matching treatment and control groups by estimated treatment propensity; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Propensity score matching.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1983) · Canonical entry
ID3 decision-tree inductionAI evaluation · Study design in development
ALG-055 · Information-gain tree learning algorithm · Proposed
Project: Compare ID3 decision-tree induction against appropriate established baselines for greedy attribute selection for classification trees; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of ID3 decision-tree induction.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1986) · Canonical entry
Lasso regressionAI evaluation · Study design in development
ALG-056 · Sparse regularised regression method · Proposed
Project: Compare Lasso regression against appropriate established baselines for l1 penalisation inducing zero coefficients; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Lasso regression.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1996) · Canonical entry
Ridge regressionAI evaluation · Study design in development
ALG-057 · L2 regularised regression method · Proposed
Project: Compare Ridge regression against appropriate established baselines for coefficient shrinkage for multicollinearity and prediction; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Ridge regression.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1970) · Canonical entry
Elastic net regressionAI evaluation · Study design in development
ALG-058 · Hybrid L1/L2 regularised regression · Proposed
Project: Compare Elastic net regression against appropriate established baselines for combining sparse selection and grouped coefficient shrinkage; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Elastic net regression.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2005) · Canonical entry
Dynamic time warpingAI evaluation · Study design in development
ALG-059 · Sequence alignment algorithm · Proposed
Project: Compare Dynamic time warping against appropriate established baselines for nonlinearly aligning time-indexed observations; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Dynamic time warping.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1978) · Canonical entry
Beam searchAI evaluation · Study design in development
ALG-060 · Heuristic sequence search · Proposed
Project: Compare Beam search against appropriate established baselines for retaining only top-scoring partial candidate paths at each step; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Beam search.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1977) · Canonical entry
Hidden Markov modelAI evaluation · Study design in development
ARC-060 · Latent-state sequential probabilistic model · Proposed
Project: Compare Hidden Markov model against appropriate established baselines for discrete hidden states producing stochastic observations; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Hidden Markov model.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1989) · Canonical entry
Conditional random fieldAI evaluation · Study design in development
ARC-061 · Discriminative structured prediction model · Proposed
Project: Compare Conditional random field against appropriate established baselines for conditional probability modelling of structured labels; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Conditional random field.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2001) · Canonical entry
Factor graphAI evaluation · Study design in development
ARC-062 · Graphical probabilistic factorisation architecture · Proposed
Project: Compare Factor graph against appropriate established baselines for representing factor-variable relationships for inference; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Factor graph.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2001) · Canonical entry
Hopfield networkAI evaluation · Study design in development
ARC-063 · Recurrent associative-memory network · Proposed
Project: Compare Hopfield network against appropriate established baselines for energy-based attractor dynamics for content-addressable recall; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Hopfield network.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (1982) · Canonical entry
Restricted Boltzmann machineAI evaluation · Study design in development
ARC-064 · Bipartite energy-based generative network · Proposed
Project: Compare Restricted Boltzmann machine against appropriate established baselines for stochastic visible and hidden units trained on observations; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Restricted Boltzmann machine.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2006) · Canonical entry
Deep belief networkAI evaluation · Study design in development
ARC-065 · Stacked probabilistic generative model · Proposed
Project: Compare Deep belief network against appropriate established baselines for hierarchies of latent representation using restricted boltzmann layers; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Deep belief network.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2006) · Canonical entry
PixelCNNAI evaluation · Study design in development
ARC-066 · Autoregressive image density architecture · Proposed
Project: Compare PixelCNN against appropriate established baselines for predicting image pixels conditioned on preceding pixels; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of PixelCNN.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2016) · Canonical entry
WaveNetAI evaluation · Study design in development
ARC-067 · Dilated causal convolutional audio generator · Proposed
Project: Compare WaveNet against appropriate established baselines for autoregressive raw-audio modelling using dilated convolutions; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of WaveNet.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2016) · Canonical entry
Transformer-XLAI evaluation · Study design in development
ARC-068 · Segment-recurrent Transformer architecture · Proposed
Project: Compare Transformer-XL against appropriate established baselines for long-context sequence modelling with cached recurrent attention states; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Transformer-XL.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2019) · Canonical entry
Capsule networkAI evaluation · Study design in development
ARC-069 · Part-whole equivariant neural architecture · Proposed
Project: Compare Capsule network against appropriate established baselines for vector capsules and routing for pose-sensitive recognition; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Capsule network.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2017) · Canonical entry
Graph TransformerAI evaluation · Study design in development
ARC-070 · Attention-based graph neural architecture family · Proposed
Project: Compare Graph Transformer against appropriate established baselines for combining graph topology and attention for graph representation learning; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Graph Transformer.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2020) · Canonical entry
Neural probabilistic language modelAI evaluation · Study design in development
ARC-071 · Distributed-embedding neural language architecture · Proposed
Project: Compare Neural probabilistic language model against appropriate established baselines for jointly learning word representations and sequence likelihood; document assumptions, independent test data, counterexamples, reproducibility and limitations.
Status: Research proposal only; implementation, evaluation and findings not completed.
Paper initiative: Review provenance, comparative performance, failure conditions and potential extensions of Neural probabilistic language model.
Unpublished initiative; no completed manuscript or results asserted.
Research reference (2003) · Canonical entry