Thermodynamic cost and benefit of memory

Abstract

This letter exposes a tight connection between the thermodynamic efficiency of information processing and predictive inference. A generalized lower bound on dissipation is derived for partially observable information engines which are allowed to use temperature differences. It is shown that the retention of irrelevant information limits efficiency. A data representation strategy is derived from optimizing a fundamental physical limit to information processing: minimizing the lower bound on dissipation leads to a data compression method that maximally retains relevant, predictive, information. In that sense, predictive inference emerges as the strategy that least precludes energy efficiency.

Accepted for publication in Physical Review Letters

References in corpus (16)

• Description of quantum coherence in thermodynamic processes requires constraints beyond free energy
• Quantum speed limits: from Heisenberg's uncertainty principle to optimal quantum control
• Second Law of Thermodynamics with Discrete Quantum Feedback Control
• High-precision test of Landauer's principle in a feedback trap
• Work and information processing in a solvable model of Maxwell's demon
• Experimental study of mutual information in a Maxwell Demon
• Quantum Szilard Engine
• Lossless Brownian information engine
• Erasure without work in an asymmetric, double-well potential
• Direct measurement of nonequilibrium system entropy is consistent with Gibbs-Shannon form
• Beyond Boltzmann-Gibbs statistics: Maximum entropy hyperensembles out-of-equilibrium
• Fundamental work cost of quantum processes
• Nanoscale virtual potentials using optical tweezers
• The underdamped Brownian duet and stochastic linear irreversible thermodynamics
• Biochemical Szilard engines for memory-limited inference
• When memory pays: Discord in hidden Markov models

Cited by in corpus (22)

• Maximizing power and velocity of an information engine
• Decomposing the local arrow of time in interacting systems
• Is stochastic thermodynamics the key to understanding the energy costs of computation?
• Information engine in a nonequilibrium bath
• Energy and information flows in autonomous systems
• Emergence of local irreversibility in complex interacting systems
• Bayesian Mechanics for Stationary Processes
• Energetic cost of feedback control
• Refining Landauer's Stack: Balancing Error and Dissipation When Erasing Information
• Unlocking the potential of information flow: Maximizing free-energy transduction in a model of an autonomous rotary molecular motor
• Optimal work extraction and mutual information in a generalized Szilárd engine
• Information Arbitrage in Bipartite Heat Engines
• Information thermodynamics: from physics to neuroscience
• Optimal Work Extraction and the Minimum Description Length Principle
• Performance limits of information engines
• Partially Observable Szilard Engines
• Finite-Time Optimization of Quantum Szilard heat engine
• Pareto-optimal clustering with the primal deterministic information bottleneck
• Rebooting Neuromorphic Hardware Design -- A Complexity Engineering Approach
• Work, entropy production, and thermodynamics of information under protocol constraints
• A Non-equilibrium Thermodynamic Framework of Consciousness
• Simple explanation of Landauer's bound and its ineffectiveness for multivalued logic