Bayesian Mechanics for Stationary Processes

Abstract

This paper develops a Bayesian mechanics for adaptive systems. Firstly, we model the interface between a system and its environment with a Markov blanket. This affords conditions under which states internal to the blanket encode information about external states. Second, we introduce dynamics and represent adaptive systems as Markov blankets at steady-state. This allows us to identify a wide class of systems whose internal states appear to infer external states, consistent with variational inference in Bayesian statistics and theoretical neuroscience. Finally, we partition the blanket into sensory and active states. It follows that active states can be seen as performing active inference and well-known forms of stochastic control (such as PID control), which are prominent formulations of adaptive behaviour in theoretical biology and engineering.

arXiv comment18 pages, 11 figures

References in corpus (28)

• Variational Inference: A Review for Statisticians
• Statistical Physics of Self-Replication
• The thermodynamics of prediction
• Stochastic Dynamical Structure (SDS) of Nonequilibrium Processes in the Absence of Detailed Balance. II: construction of SDS with nonlinear force and multiplicative noise
• A free energy principle for a particular physics
• A Complete Recipe for Stochastic Gradient MCMC
• Reconstruction of Gaussian quantum mechanics from Liouville mechanics with an epistemic restriction
• The Zig-Zag Process and Super-Efficient Sampling for Bayesian Analysis of Big Data
• Optimal non-reversible linear drift for the convergence to equilibrium of a diffusion
• Improved bounds on entropy production in living systems
• Convergence of Numerical Time-Averaging and Stationary Measures via Poisson Equations
• Deep Active Inference
• A Novel Adaptive Controller for Robot Manipulators based on Active Inference
• Irreversible Langevin samplers and variance reduction: a large deviation approach
• Active inference body perception and action for humanoid robots
• End-to-End Pixel-Based Deep Active Inference for Body Perception and Action
• A piecewise deterministic scaling limit of Lifted Metropolis-Hastings in the Curie-Weiss model
• Some interesting observations on the free energy principle
• A Decomposition of Irreversible Diffusion Processes Without Detailed Balance
• Neural Stochastic Differential Equations: Deep Latent Gaussian Models in the Diffusion Limit
• Learning and inference in a nonequilibrium Ising model with hidden nodes
• Thermodynamic cost and benefit of memory
• Uncertainty relations and fluctuation theorems for Bayes nets
• Marginal and Conditional Second Laws of Thermodynamics
• How particular is the physics of the free energy principle?
• A Unifying and Canonical Description of Measure-Preserving Diffusions
• Thermodynamic Uncertainty Relations for Multipartite Processes
• On the thermodynamics of prediction under dissipative adaptation

Cited by in corpus (8)

• The free energy principle made simpler but not too simple
• Path integrals, particular kinds, and strange things
• Some interesting observations on the free energy principle
• A variational synthesis of evolutionary and developmental dynamics
• Cell decision-making through the lens of Bayesian learning
• Geometric Methods for Sampling, Optimisation, Inference and Adaptive Agents
• Kalman filters as the steady-state solution of gradient descent on variational free energy
• A Mathematical Walkthrough and Discussion of the Free Energy Principle