bhairavmehta95 / learning-relevant-tensor-networks

Implementation of Learning Relevant Features from Multi-Scale Tensor Networks

learning-relevent-tensor-networks

We provide an efficient, multiprocess implementation of the algorithm proposed in Stoudenmires 2018.

How to use it

The algorithm contains two phases:

1. Unsupervised coarse graining: This allows to construct the tree tensor layer U.

python -m experiments.driver_ucg --{arg1}={val1} ...

2. Supervised optimization of the top tensor: Once we have computed U, we just need to optimize the top tensor.

python -m experiments.driver_train_top --{arg1}={val1} ...