[S18] Entropy from Machine Learning - R. Janik скачать в хорошем качестве

[S18] Entropy from Machine Learning - R. Janik

Romauld A. Janik Institute of Physics, Jagiellonian University Abstract: In this talk I will describe a Machine Learning based method which allows to compute the entropy (and equivalently the free energy) directly from a set of Monte Carlo configurations at a given temperature. The problem of calculating the entropy of a set of binary configurations/signals is translated into a sequence of supervised classification tasks. Subsequently, one can use virtually any machine learning classification algorithm for computing entropy. Other potential applications include computing the entropy of spiking neurons or any other multidimensional binary signals.​