dingliu0305 / Hybrid-Tensor-Network

Welcome to star this projects, it will be helpful for you to learn tensor network caculating in Pytorch.

This is the computational appendix for the following paper:

Ding Liu, Zekun Yao, Quan Zhang. Quantum-Classical Machine learning by Hybrid Tensor Networks. arXiv:2005.09428v1, 2020.

Files

• HTN_MNIST：The baseline for fashion-MNIST dataset.If you want to train the MNIST dataset, you can just modify the torchvision.datasets.FashionMNIST to torchvision.datasets.MNIST。The best HTN result 90% on fashion-MNIST will come true.And you can also just load the model from great_checkpoint/fashionmnist-90%ACC_checkpoint.pth to test.
• HTN_MNIST_V2：Generating loss curve figure and Adding tensorboard to monitor the loss and acc.
• HTN_MNIST_CNN_V3：Modifing the Neural network structure with CNN.
• HTN_MNIST_eval：Loading the .pth type file to evaluate the fashion-MNIST(or MNIST)，and you will get an accuracy about testing.
• HTN_AutoEncoder.py：Use the training model from.pth file to encode the images.