v15hv4/LearnEntropy

This code estimates the entropy production rate by machine learning of trajectory data. The method is based on the thermodynamic uncertainty relation.

Learning estimators for the entropy production

1. About

You can estimate the entropy production rate and thermodynamic force from trajectory data. This is a rearranged version of the code used in the following papers:

• Otsubo, Ito, Dechant, and Sagawa, Phys. Rev. E 101, 062106, 2020.
• Otsubo, Sreekanth, Sagawa, and Krishnamurthy, arXiv:2010.03852, 2020.

2. How to use

The code is written with Python3, and depends on the following libraries.

• torch, numpy, matplotlib, pillow

Please see demo_stationary.ipynb and demo_nonstationary.ipynb for the usage. Their contents can also be checked by reading the corresponding HTML files.