Overview
This document is a call for participation in the initiative "Towards Active Diffusion: A Tale of Multiple (den)Cities" (2023-AD). The work will characterize mathematical formalisms and computational applications of Active Inference and Diffusion Models, and explore relevant applications.
Overview source: Curated paper metadata.
Methods and contributions
Read the source for the full argument, qualifications, and evidence.
Findings and contributions
- The document states the 2023-AD mission: characterizing mathematical formalisms and computational applications of Active Inference and Diffusion Models.
- It identifies shared ground between ActInf and DMs, including that both draw on stochastic thermodynamics for their internal representations.
- It proposes that LDMs' compression of high-dimensional input into latent representations is analogous to how ActInf agents encode beliefs about their environment.
Methods
- Proposed theoretical analysis of formal relations between ActInf and diffusion models — The catechism plans (for 2023) an analytical paper and literature review comparing Latent Diffusion Models with belief propagation in Active Inference.
- Proposed integration of diffusion models into the Active Inference formalism — Planned applied work: build on existing implementations and design experiments testing DMs as learned representations within the action-perception loop.
- Proposed cadCAD-based multi-agent work extending Active Blockference — Plans to use the cadCAD complex adaptive dynamics package for cognitive ecosystem design and to extend the Active Blockference package.
Summary sources: Paper metadata and evidence · Extracted source text.
Document downloads
PDFs are hosted here; supporting sources may use pinned archives.
- 2022_ActiveDiffusion.pdf PDF · 0.10 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.