Gibbs entropy over measurement vectors is claimed to track neural-network compression performance, and a compressed-sensing pruning method (DTC) is introduced as supporting evidence.
Understanding entropy production via a thermal zero-player game
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
Understanding the natural bounds of entropy production for driven nonequilibrium dynamics in many-body systems reveals how the fundamentals of thermodynamics manifest in these regimes across a wide variety of systems. In this direction, we propose and study the dynamics of a thermal zero-player entropy game, the Ising-Conway Entropy Game (ICEg), a self-driven system exhibiting characteristics of lattice gases, Ising models, and discrete games. We show that there is a universal bound on the entropy production rate, independent of temperature and lattice size. The thermalized game is shown to be physically interesting and a plausible testbed for studying the fundamentals of stochastic thermodynamics.
citation-role summary
citation-polarity summary
fields
stat.ML 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Gibbs randomness-compression proposition
Gibbs entropy over measurement vectors is claimed to track neural-network compression performance, and a compressed-sensing pruning method (DTC) is introduced as supporting evidence.