arXiv:1909.10831

[refraction-ray]

• Link
• Title: Entropy from Machine Learning
• Keywords (optional):
• Authors (optional):
• Reason (optional):
• Summary (optional): Estimate entropy from MC sample data for 2D Ising model using typical ML classifier. Eq 15 is the essence, but not sure about the training object for p, why minimize it? Some might-be understanding: this approach seems to always overestimate the entropy, since independence of all binaries give the maximal possible entropy. So minimizing the loss, is just trying to find dependence of x_n on x_1 to x_n-1. It is interesting that all results presented in this work overestimates the real entropy. See the proof on left side of page4
• Ideas (optional):
  • Data augmentation: In the context of the Ising model, we can rotate the spin configurations by multiples of 90◦ or perform arbitrary (periodic) translations. Symmetries for bigger data!

[refraction-ray]

In practical, no idea whether this method is good and efficient enough for practical model entropy estimation. If this is the case, the method presented in this work would definitely be very important! And also a very interesting application of ML ingredients on computational physics.

[refraction-ray]

Entropy in MC from old days

• Entropy of spin models by the Monte Carlo method (1995)
• Entanglement of Interacting Fermions in Quantum Monte Carlo (2013)
• Measuring Renyi Entanglement Entropy in Quantum Monte Carlo Simulations