Thermodynamic Regulation of Finite-Time Gibbs Training in Energy-Based Models: A Restricted Boltzmann Machine Study

• URL: http://arxiv.org/abs/2603.02525v1
• Date: Tue, 03 Mar 2026 02:21:19 GMT
• Title: Thermodynamic Regulation of Finite-Time Gibbs Training in Energy-Based Models: A Restricted Boltzmann Machine Study
• Authors: Görkem Can Süleymanoğlu,
• Abstract summary: We introduce an endogenous thermodynamic regulation in which temperature evolves as a state variable to sampling statistics.<n>Experiments on MNIST show that the proposed self-regulated approximation regime preserves stability and effective sample performance.<n>Overall, the results show that WeM training as a controlled non-equilibrium process rather than a static equilibrium process.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Restricted Boltzmann Machines (RBMs) are typically trained using finite-length Gibbs chains under a fixed sampling temperature. This practice implicitly assumes that the stochastic regime remains valid as the energy landscape evolves during learning. We argue that this assumption can become structurally fragile under finite-time training dynamics. This fragility arises because, in nonconvex energy-based models, fixed-temperature finite-time training can generate admissible trajectories with effective-field amplification and conductance collapse. As a result, the Gibbs sampler may asymptotically freeze, the negative phase may localize, and, without sufficiently strong regularization, parameters may exhibit deterministic linear drift. To address this instability, we introduce an endogenous thermodynamic regulation framework in which temperature evolves as a dynamical state variable coupled to measurable sampling statistics. Under standard local Lipschitz conditions and a two-time-scale separation regime, we establish global parameter boundedness under strictly positive L2 regularization. We further prove local exponential stability of the thermodynamic subsystem and show that the regulated regime mitigates inverse-temperature blow-up and freezing-induced degeneracy within a forward-invariant neighborhood. Experiments on MNIST demonstrate that the proposed self-regulated RBM substantially improves normalization stability and effective sample size relative to fixed-temperature baselines, while preserving reconstruction performance. Overall, the results reinterpret RBM training as a controlled non-equilibrium dynamical process rather than a static equilibrium approximation.

Related papers

• Avoiding Premature Collapse: Adaptive Annealing for Entropy-Regularized Structural Inference [1.7523718031184992]
  We identify a fundamental mechanism for this failure: textbfPremature Mode Collapse.<n>We propose textbfEfficient Piecewise Hybrid Adaptive Stability Control (EPH-ASC), an adaptive scheduling algorithm that monitors the stability of the inference process.
  arXiv  Detail & Related papers  (2026-01-30T14:47:18Z)
• Stability as a Liability:Systematic Breakdown of Linguistic Structure in LLMs [5.96875296117642]
  We show that stable parameter trajectories lead stationary solutions to minimize the forward KL divergence to the empirical distribution.<n>We empirically validate this effect using a controlled feedback-based training framework.<n>It indicates that optimization stability and generative expressivity are not inherently aligned, and that stability alone is an insufficient indicator of generative quality.
  arXiv  Detail & Related papers  (2026-01-26T15:34:50Z)
• Temperature dependence of energy transport in the $\mathbb{Z}_3$ chiral clock model [0.0]
  We study energy transport within the non-integrable regime of the one-dimensional $mathbbZ_3$ chiral clock model. We extract the transport coefficients of the model at relatively high temperatures above both its gapless and gapped low-temperature phases. Although we are not yet able to reach temperatures where quantum critical scaling would be observed, our approach is able to access the transport properties of the model.
  arXiv  Detail & Related papers  (2023-10-31T18:00:30Z)
• Characterizing the spontaneous collapse of a wavefunction through entropy production [0.0]
  We investigate the phenomenology leading to the non-conservation of energy of the continuous spontaneous localization (CSL) model. We use such framework to assess the equilibration process entailed by the dissipative formulation of the model (dCSL) We show that the CSL model violates Clausius law, as it exhibits a negative entropy production rate.
  arXiv  Detail & Related papers  (2023-10-19T12:10:52Z)
• Machine learning in and out of equilibrium [58.88325379746631]
  Our study uses a Fokker-Planck approach, adapted from statistical physics, to explore these parallels. We focus in particular on the stationary state of the system in the long-time limit, which in conventional SGD is out of equilibrium. We propose a new variation of Langevin dynamics (SGLD) that harnesses without replacement minibatching.
  arXiv  Detail & Related papers  (2023-06-06T09:12:49Z)
• Apparent pathologies in stochastic entropy production in the thermalisation of an open two-level quantum system [0.0]
  We investigate the entropic consequences of the relaxation of an open two-level quantum system towards a thermalised statistical state. We demonstrate that thermalisation starting from a general state is accompanied by a persistent non-zero mean rate of change of the environmental component of entropy production.
  arXiv  Detail & Related papers  (2023-03-07T11:34:46Z)
• Clean two-dimensional Floquet time-crystal [68.8204255655161]
  We consider the two-dimensional quantum Ising model, in absence of disorder, subject to periodic imperfect global spin flips. We show by a combination of exact diagonalization and tensor-network methods that the system can sustain a spontaneously broken discrete time-translation symmetry. We observe a non-perturbative change in the decay rate of the order parameter, which is related to the long-lived stability of the magnetic domains in 2D.
  arXiv  Detail & Related papers  (2022-05-10T13:04:43Z)
• Observation of Time-Crystalline Eigenstate Order on a Quantum Processor [80.17270167652622]
  Quantum-body systems display rich phase structure in their low-temperature equilibrium states. We experimentally observe an eigenstate-ordered DTC on superconducting qubits. Results establish a scalable approach to study non-equilibrium phases of matter on current quantum processors.
  arXiv  Detail & Related papers  (2021-07-28T18:00:03Z)
• Uhlmann Fidelity and Fidelity Susceptibility for Integrable Spin Chains at Finite Temperature: Exact Results [68.8204255655161]
  We show that the proper inclusion of the odd parity subspace leads to the enhancement of maximal fidelity susceptibility in the intermediate range of temperatures. The correct low-temperature behavior is captured by an approximation involving the two lowest many-body energy eigenstates.
  arXiv  Detail & Related papers  (2021-05-11T14:08:02Z)
• Analog cosmological reheating in an ultracold Bose gas [58.720142291102135]
  We quantum-simulate the reheating-like dynamics of a generic cosmological single-field model in an ultracold Bose gas. Expanding spacetime as well as the background oscillating inflaton field are mimicked in the non-relativistic limit. The proposed experiment has the potential of exploring the evolution up to late times even beyond the weak coupling regime.
  arXiv  Detail & Related papers  (2020-08-05T18:00:26Z)
• Probing eigenstate thermalization in quantum simulators via fluctuation-dissipation relations [77.34726150561087]
  The eigenstate thermalization hypothesis (ETH) offers a universal mechanism for the approach to equilibrium of closed quantum many-body systems. Here, we propose a theory-independent route to probe the full ETH in quantum simulators by observing the emergence of fluctuation-dissipation relations. Our work presents a theory-independent way to characterize thermalization in quantum simulators and paves the way to quantum simulate condensed matter pump-probe experiments.
  arXiv  Detail & Related papers  (2020-07-20T18:00:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.