Machine learning algorithms based on generalized Gibbs ensembles

Abstract

Machine learning algorithms often take inspiration from the established results and knowledge from statistical physics. A prototypical example is the Boltzmann machine algorithm for supervised learning, which utilizes knowledge of classical thermal partition functions and the Boltzmann distribution. Recently, a quantum version of the Boltzmann machine was introduced by Amin, et. al., however, noncommutativity of quantum operators renders the training process by minimizing a cost function inefficient. Recent advances in the study of non-equilibrium quantum integrable systems, which never thermalize, have lead to the exploration of a wider class of statistical ensembles. These systems may be described by the so-called generalized Gibbs ensemble (GGE), which incorporates a number of "effective temperatures". We propose that these GGEs can be successfully applied as the basis of a Boltzmann-machine-like learning algorithm, which operates by learning the optimal values of effective temperatures. We show that the GGE algorithm is an optimal quantum Boltzmann machine: it is the only quantum machine that circumvents the quantum training-process problem. We apply a simplified version of the GGE algorithm, where quantum effects are suppressed, to the classification of handwritten digits in the MNIST database. While lower error rates can be found with other state-of-the-art algorithms, we find that our algorithm reaches relatively low error rates while learning a much smaller number of parameters than would be needed in a traditional Boltzmann machine, thereby reducing computational cost. * t.puskarov@uu.nl † a.cortescubero@uu.nl arXiv:1804.03546v3 [cond-mat.stat-mech] 1 Dec 2018

Recent Advances for Quantum Neural Networks in Generative Learning

“…Puškarov and Cubero [219] argued that the generalized Gibbs ensembles [220,221] could be leveraged as a basis for QBMs. In doing so, the gradient information can be effectively obtained due to the commutativity of the conserved charge in the integrable Hamiltonian by learning the optimal effective temperatures.…”

An Enquiry on similarities between Renormalization Group and Auto-Encoders using Transfer Learning

“…This is where ML can be of immense assistance [7][8][9][10][11]. On the other hand, physicists have also been contributing to ML, drawing inspiration from the theories and techniques developed for computational physics, as well as providing insights into the foundation of AI and ML [12][13][14][15]. One such insight manifests as we study deep learning side by side with the renormalization group (Fig.…”