Using machine learning to characterize 2d Ising magnetization states

The purpose of this project is to independently validate the claim made by Machine Learning Phases of Matter (1605.01735). It is a neural net built with Tensorflow to identify magnetization states of 2d Ising lattices at low temperatures around the critical temperature.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. Build passing with python 3.5.2.

Prerequisites

Dependencies will be handled by pip, make sure you have python3 installed at least version 3.5

pip install -r requirements.txt

Ensure you have enough memory before running the code, the code is not yet optimized for memory allocation, could take up 1-2gb of memory.

Running the code

Run the code with python

python tf.py

Your data should begin generating, or if you have the pre-generated data.p file, tensorflow will build the model and begin training. The model is expected to have ~98% accuracy on the validation set, and similar accuracy in the test set if you have it loaded.

Built With

• Tensorflow - To build the neural net

Authors

• Danny Kong - Github

Acknowledgments

• This code was inspired by the paper Machine Learning Phases of Matter (Carrasquilla, J; Melko, R)
• Metropolis algorithm was implemented by Rajesh Rinet Github, our data was generated by running his algorithm, equilibrating, and sampling at discrete time points