Overview
This paper introduces Generalized Notation Notation (GNN), a novel approach to generative model representation that facilitates communication, understanding, and application of Active Inference across various domains. GNN complements the Active Inference Ontology as a flexible and expressive language for education and modeling, by providing a standardized method for describing cognitive models. In this paper we introduce GNN, and provide a Step-by-Step example of what GNN looks like in practice. We then explore "the Triple Play", a pragmatic approach to expressing GNN in linguistic, visual, and executable cognitive models. By situating GNN within the broader context of cognitive modeling and Active Inference, this work aims to bridge and respect the gaps among different modeling settings. The goal of this work is to facilitate interdisciplinary research and application, ultimately promoting the advancement of the field. Github: https://github.com/ActiveInferenceInstitute/GeneralizedNotationNotation Coda: https://coda.io/@active-inference-institute/generalized-notation-notation
Overview source: Curated paper metadata.
Methods and contributions
Read the source for the full argument, qualifications, and evidence.
Findings and contributions
- Introduces GNN as a standardized method for describing cognitive models that complements the Active Inference Ontology for education and modeling.
- Argues GNN's plain-text basis allows rendering into mathematical notation, figures, natural language, pseudocode and executable simulations, and use with regular expressions and LLMs.
- As pseudocode, GNN does not restrict which programming language or package implements a model, which the authors argue aids backward and forward compatibility.
- Names future directions: better automatic rendering software, tighter Active Inference Ontology integration, and cadCAD for execution order and parameter sweeps.
Methods
- GNN specification: ASCII model description in Markdown-structured source files — Defines Generalized Notation Notation, which writes Active Inference generative models with ASCII letters and punctuation in a Markdown-style source file.
- Worked example re-expressing the Smith, Friston & Whyte 2022 step-by-step tutorial — Demonstrates GNN by following the same progressive model-building sequence as the 2022 step-by-step Active Inference tutorial.
- 'Triple Play' framing: text, graphical, and executable model renderings — Proposes expressing a GNN model across three modalities: text-based models, statistical graphical models, and executable cognitive models.
Summary sources: Paper metadata and evidence · Extracted source text.
Document downloads
PDFs are hosted here; supporting sources may use pinned archives.
- 2023_GNN.pdf PDF · 0.75 MiB · extracted-text source
Citation
Citation metadata follows the unified bibliography.
Catalog details and resources
Related works
Ranked by recorded methods, topics, contributors and associated software, with contributor roles and domain-only matches labelled. Metadata overlap is not a citation or a scientific equivalence.