TF TCN

Tensorflow Temporal Convolutional Network

This is an implementation of An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling in TensorFlow.

I've verified that given same argument, my network has exactly same number of parameter as his model. It is able to reach the same loss/accuracy level in these problems, BUT sometimes it gets good result a little slower than the original implementation in Torch.

This repository mainly follows the structure of the original repo. And for illustration of different tasks, you could take a look at keras TCN. The author provides some nice figures there.

Some codes are modified from original implementation, keras TCN, and openai.

Domains and Datasets

This repository contains the benchmarks to the following tasks, with details explained in each sub-directory:

• The Adding Problem with various T (we evaluated on T=200, 400, 600)

• Copying Memory Task with various T (we evaluated on T=500, 1000, 2000)

• Sequential MNIST digit classification

• Permuted Sequential MNIST (based on Seq. MNIST, but more challenging)

• PennTreebank [SMALL] word-level language modeling (LM)

Run

In the root directory of this repo, type the following to run different experiments

python3 -m [module_name] [args]
python3 -m adding_problem.add_test [args]
python3 -m copymem.copymem_test [args]
python3 -m mnist_pixel.pmnist_test.py --epo 10

References

[1] Bai, Shaojie, J. Zico Kolter, and Vladlen Koltun. "An empirical evaluation of generic convolutional and recurrent networks for sequence modeling." arXiv preprint arXiv:1803.01271 (2018). [2] Salimans, Tim, and Diederik P. Kingma. "Weight normalization: A simple reparameterization to accelerate training of deep neural networks." Advances in Neural Information Processing Systems. 2016.