Inferring in Circle: Active Inference in Continuous State Space Using Hierarchical Gaussian Filtering of Sufficient Statistics

Peter Thestrup Waade*, Nace Mikus, Chris Mathys

*Corresponding author for this work

Research output: Contribution to book/anthology/report/proceedingArticle in proceedingsResearchpeer-review

Abstract

We create a continuous state space active inference agent based on the hierarchical Gaussian filter. It uses the HGF to track the sufficient statistics of noisy observations of a moving target that is performing a Gaussian random walk with drift and varying volatility. On the basis of this filtering, the agent predicts the target’s position, and minimizes surprisal by staying close to it. Our simulated agent represents the first full implementation of this approach. It demonstrates the feasibility of supplementing active inference with HGF-filtering of the sufficient statistics of observations, which is particularly useful in noisy and volatile continuous state space environments.

Original languageEnglish
Title of host publicationMachine Learning and Principles and Practice of Knowledge Discovery in Databases: International Workshops of ECML PKDD 2021, Proceedings
Number of pages9
Volume1
PublisherSpringer
Publication dateFeb 2022
Pages810-818
ISBN (Print)9783030937355
DOIs
Publication statusPublished - Feb 2022
Event21st European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2021 - Virtual, Online
Duration: 13 Sept 202117 Sept 2021

Conference

Conference21st European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2021
LocationVirtual, Online
Period13/09/202117/09/2021
SeriesCommunications in Computer and Information Science
Number1524
ISSN1865-0929

Keywords

  • Active inference
  • Continuous state space
  • Hierarchical gaussian filter
  • Precision-weighted prediction errors
  • Sufficient statistics filtering

Fingerprint

Dive into the research topics of 'Inferring in Circle: Active Inference in Continuous State Space Using Hierarchical Gaussian Filtering of Sufficient Statistics'. Together they form a unique fingerprint.

Cite this