FR-01 · Typed relation and context graph
Across most families: subject, relation, object, polarity, domain, context and conditions. A graph records claims but does not prove them.
Canonical taxonomy updated: This is a historical research view. The current four-level classification separates AI categories, learning paradigms, methods, and model/system architectures; retired labels are traceable without being counted as distinct active types. Open current taxonomy (v1.2) · Matching CSV.
Research & Knowledge · AI and machine-learning development
One unique architecture family, model class or technical method per row. Every listed type has its own proposed project and research paper initiative. Related methods and broader learning paradigms are classified rather than counted repeatedly.
Current evidence-based register: 40 distinct base families or methods. This is not a claim that only 40 AI types exist. The previous set of 1,000 entries represented 40 families × 25 proposed scenarios, not 1,000 separate types. Expansion requires individually identified, nonduplicate types and credible source literature.
Taxonomy revision: AI application categories, model and system architectures, and statistical or machine-learning methods are now separated in three aligned tables. Survival analysis is classified as a statistical method, not an AI architecture. Open the three corrected taxonomy tables.
Each numbered row is unique. A project or paper link identifies a planned initiative, not a completed build or publication. Wikipedia is introductory context, not independent scientific validation.
| # | AI type / method | Classification | Category | Wikipedia | Manifold Grace project | Paper initiative | Status |
|---|---|---|---|---|---|---|---|
| 1 | CNN feature hierarchy | model architecture | Computer vision | Wikipedia | Project AIT-0001 | Paper initiative | Proposed |
| 2 | Vision Transformer | model architecture | Computer vision | Wikipedia | Project AIT-0002 | Paper initiative | Proposed |
| 3 | Object detection network | model architecture class | Computer vision | Wikipedia | Project AIT-0003 | Paper initiative | Proposed |
| 4 | Semantic segmentation network | model architecture class | Computer vision | Wikipedia | Project AIT-0004 | Paper initiative | Proposed |
| 5 | Instance segmentation network | model architecture class | Computer vision | Wikipedia | Project AIT-0005 | Paper initiative | Proposed |
| 6 | Encoder Transformer | model architecture | Language and knowledge | Wikipedia | Project AIT-0006 | Paper initiative | Proposed |
| 7 | Autoregressive Transformer | model architecture | Language and knowledge | Wikipedia | Project AIT-0007 | Paper initiative | Proposed |
| 8 | Retrieval-augmented generation | system integration pattern | Language and knowledge | Wikipedia | Project AIT-0008 | Paper initiative | Proposed |
| 9 | Sequence-to-sequence Transformer | model architecture | Language and knowledge | Wikipedia | Project AIT-0009 | Paper initiative | Proposed |
| 10 | Knowledge-graph question answering | application architecture | Language and knowledge | Wikipedia | Project AIT-0010 | Paper initiative | Proposed |
| 11 | Gradient-boosted trees | machine-learning algorithm family | Predictive and tabular | Wikipedia | Project AIT-0011 | Paper initiative | Proposed |
| 12 | Generalised additive model | statistical model family | Predictive and tabular | Wikipedia | Project AIT-0012 | Paper initiative | Proposed |
| 13 | Temporal forecasting Transformer | model architecture | Predictive and tabular | Wikipedia | Project AIT-0013 | Paper initiative | Proposed |
| 14 | State-space sequence model | model architecture | Predictive and tabular | Wikipedia | Project AIT-0014 | Paper initiative | Proposed |
| 15 | Survival and hazard model | statistical model family | Predictive and tabular | Wikipedia | Project AIT-0015 | Paper initiative | Proposed |
| 16 | Graph convolutional network | model architecture | Graphs and relational data | Wikipedia | Project AIT-0016 | Paper initiative | Proposed |
| 17 | Graph attention network | model architecture | Graphs and relational data | Wikipedia | Project AIT-0017 | Paper initiative | Proposed |
| 18 | Relational graph neural network | model architecture | Graphs and relational data | Wikipedia | Project AIT-0018 | Paper initiative | Proposed |
| 19 | Temporal graph neural network | model architecture | Graphs and relational data | Wikipedia | Project AIT-0019 | Paper initiative | Proposed |
| 20 | Neural relational database model | research model/system concept | Graphs and relational data | Wikipedia | Project AIT-0020 | Paper initiative | Proposed |
| 21 | Denoising diffusion model | generative model architecture | Generative and multimodal | Wikipedia | Project AIT-0021 | Paper initiative | Proposed |
| 22 | Latent diffusion model | generative model architecture | Generative and multimodal | Wikipedia | Project AIT-0022 | Paper initiative | Proposed |
| 23 | Generative adversarial network | generative model architecture | Generative and multimodal | Wikipedia | Project AIT-0023 | Paper initiative | Proposed |
| 24 | Variational autoencoder | generative model architecture | Generative and multimodal | Wikipedia | Project AIT-0024 | Paper initiative | Proposed |
| 25 | Vision-language model | multimodal model architecture class | Generative and multimodal | Wikipedia | Project AIT-0025 | Paper initiative | Proposed |
| 26 | Tool-using LLM agent | agent system architecture | Agents and automation | Wikipedia | Project AIT-0026 | Paper initiative | Proposed |
| 27 | Plan-and-execute agent | agent orchestration pattern | Agents and automation | Wikipedia | Project AIT-0027 | Paper initiative | Proposed |
| 28 | Multi-agent orchestration | multi-agent system pattern | Agents and automation | Wikipedia | Project AIT-0028 | Paper initiative | Proposed |
| 29 | AI-enabled BPM workflow | business workflow architecture | Agents and automation | Wikipedia | Project AIT-0029 | Paper initiative | Proposed |
| 30 | Event-driven decision agent | agent system architecture | Agents and automation | Wikipedia | Project AIT-0030 | Paper initiative | Proposed |
| 31 | Vision-based robotic policy | robotics policy architecture | Robotics and control | Wikipedia | Project AIT-0031 | Paper initiative | Proposed |
| 32 | Model-predictive controller | control method | Robotics and control | Wikipedia | Project AIT-0032 | Paper initiative | Proposed |
| 33 | Reinforcement learning policy | learning-policy method | Robotics and control | Wikipedia | Project AIT-0033 | Paper initiative | Proposed |
| 34 | Imitation learning policy | learning-policy method | Robotics and control | Wikipedia | Project AIT-0034 | Paper initiative | Proposed |
| 35 | Spiking neural controller | neural control architecture | Robotics and control | Wikipedia | Project AIT-0035 | Paper initiative | Proposed |
| 36 | Rule-based expert system | symbolic AI system | Logic, search and hybrid decision | Wikipedia | Project AIT-0036 | Paper initiative | Proposed |
| 37 | Constraint-satisfaction solver | constraint-solving algorithm | Logic, search and hybrid decision | Wikipedia | Project AIT-0037 | Paper initiative | Proposed |
| 38 | Bayesian network | probabilistic graphical model | Logic, search and hybrid decision | Wikipedia | Project AIT-0038 | Paper initiative | Proposed |
| 39 | Neuro-symbolic reasoning system | hybrid reasoning architecture | Logic, search and hybrid decision | Wikipedia | Project AIT-0039 | Paper initiative | Proposed |
| 40 | Case-based reasoning system | reasoning method | Logic, search and hybrid decision | Wikipedia | Project AIT-0040 | Paper initiative | Proposed |
Computer vision and natural-language processing are application domains; supervised and unsupervised learning are training paradigms; neural networks and convolutional networks are related model families; robotic process automation is a workflow technology; and blockchain is an infrastructure technology. They belong in the wider AI projects taxonomy, but should not all be labelled model architectures. Learning paradigms reference.
The historical 1,000-row combinatorial research dataset is preserved separately with its existing identifiers. It must not be used to inflate the number of unique architectures.
Each of the 40 base families has its own linked Wikipedia context, native relationship type and specialised component set. These are implementation candidates rather than verified deployments.
| Category | Wikipedia-linked family | Native relationship | Specialised components |
|---|---|---|---|
| Computer vision | CNN feature hierarchy | local spatial receptive fields | convolution backbone; spatial feature pyramid; image calibration |
| Computer vision | Vision Transformer | patch-token global attention | patch encoder; attention maps; positional geometry |
| Computer vision | Object detection network | object localisation and identity | detector head; box matching; non-max suppression |
| Computer vision | Semantic segmentation network | dense pixel-level category assignment | mask decoder; pixel alignment; boundary checks |
| Computer vision | Instance segmentation network | separate masks for object instances | instance tracking; mask matching; overlap verification |
| Language and knowledge | Encoder Transformer | contextual bidirectional language representation | tokeniser; encoder; entailment head |
| Language and knowledge | Autoregressive Transformer | next-token generation with long-context dependencies | decoder; context window management; factuality evaluator |
| Language and knowledge | Retrieval-augmented generation | retrieval-to-answer grounding relationships | retriever; reranker; citation entailment checker |
| Language and knowledge | Sequence-to-sequence Transformer | input-output sequence transformations | encoder-decoder; alignment tracer; translation evaluator |
| Language and knowledge | Knowledge-graph question answering | entity-link and graph-path evidence | entity linker; SPARQL query builder; path verifier |
| Predictive and tabular | Gradient-boosted trees | nonlinear feature-outcome interactions | tabular preprocessing; tree ensemble; feature attribution |
| Predictive and tabular | Generalised additive model | additive, interpretable predictors | spline feature functions; calibration plots; residual checks |
| Predictive and tabular | Temporal forecasting Transformer | time-varying series dependencies | window sampler; horizon evaluator; temporal attention |
| Predictive and tabular | State-space sequence model | latent evolving state relationships | state update kernel; filter; state diagnostics |
| Predictive and tabular | Survival and hazard model | event-time and censoring relationships | censoring handler; hazard estimation; calibration |
| Graphs and relational data | Graph convolutional network | neighbourhood topology aggregation | graph encoder; adjacency manager; graph split evaluator |
| Graphs and relational data | Graph attention network | attention-weighted neighbourhood relationships | attention layers; edge sampler; neighbourhood audit |
| Graphs and relational data | Relational graph neural network | typed edges and heterogeneous relations | relation-aware message passing; edge type validator |
| Graphs and relational data | Temporal graph neural network | time-stamped dynamic graph changes | event buffer; temporal neighbour sampler; leakage checks |
| Graphs and relational data | Neural relational database model | foreign-key and multi-table entity relations | schema parser; table join graph; relational encoder |
| Generative and multimodal | Denoising diffusion model | noise-to-sample denoising trajectories | noise scheduler; denoiser; output consistency checks |
| Generative and multimodal | Latent diffusion model | compressed latent-to-output relationships | latent encoder; denoiser; decoder; reconstruction checks |
| Generative and multimodal | Generative adversarial network | generator-discriminator training dynamics | generator; discriminator; mode-collapse evaluator |
| Generative and multimodal | Variational autoencoder | latent distribution and reconstruction relations | probabilistic encoder; latent sampler; decoder |
| Generative and multimodal | Vision-language model | cross-modal image-text correspondences | visual encoder; text encoder; alignment verifier |
| Agents and automation | Tool-using LLM agent | model-to-tool requests and permissions | tool registry; approval gate; audit trail |
| Agents and automation | Plan-and-execute agent | goal decomposition and action dependencies | planner; executor; plan validity checker |
| Agents and automation | Multi-agent orchestration | delegation and message provenance | agent router; role policy; inter-agent trace |
| Agents and automation | AI-enabled BPM workflow | tasks, transitions and business rules | BPMN engine; rule adapter; process conformance checker |
| Agents and automation | Event-driven decision agent | event conditions and triggered actions | event bus; decision table; idempotency and rollback |
| Robotics and control | Vision-based robotic policy | visual input to motor action | camera perception; policy controller; safety envelope |
| Robotics and control | Model-predictive controller | future-state planning under constraints | dynamics model; trajectory optimiser; constraint monitor |
| Robotics and control | Reinforcement learning policy | state-action-reward dynamics | policy/value networks; simulator; reward audit |
| Robotics and control | Imitation learning policy | demonstration-to-action generalisation | trajectory dataset; imitation learner; shift monitor |
| Robotics and control | Spiking neural controller | event-driven neural timing dynamics | spiking units; temporal encoder; event simulator |
| Logic, search and hybrid decision | Rule-based expert system | if-then inferences and exception rules | rule engine; conflict resolver; justification trace |
| Logic, search and hybrid decision | Constraint-satisfaction solver | variables, constraints and feasible assignments | constraint model; solver; proof-of-infeasibility reporter |
| Logic, search and hybrid decision | Bayesian network | conditional probabilistic dependence | DAG model; inference engine; posterior calibration |
| Logic, search and hybrid decision | Neuro-symbolic reasoning system | learned representations with explicit symbolic rules | neural encoder; logic layer; contradiction detector |
| Logic, search and hybrid decision | Case-based reasoning system | analogy and difference across past cases | case memory; similarity measure; adaptation validator |
The first table prioritises legibility. A second-stage expansion should add: input modality; training objective; supervision mode; parameterisation; inference mechanism; memory/state representation; output modality; deployment context; edge/cloud constraints; data provenance; intended use; limitations; model/dataset versions; evaluation dataset; result metrics; uncertainty; safety controls; legal applicability; system owner; evidence links; reviewer and reassessment trigger. Do not fill fields with guesses or treat missing research as completed.
Keep numerical performance and compliance claims separate from architecture categories. A research candidate without a prototype must have an explicit not tested status.
Wikipedia links provide reader context only; research validity and architecture naming require independent technical literature. Proposed hybrids are not automatically distinct established subfields, nor is business process automation itself necessarily a standalone neural-network architecture.