Temporal Convolutional Networks (TCNs) are promising Deep Learning models for time-series processing tasks. One key feature of TCNs is time-dilated convolution, whose optimization requires extensive experimentation. We propose an automatic dilation optimizer, which tackles the problem as a weight pruning on the time-axis, and learns dilation factors together with weights, in a single training. Our method reduces the model size and inference latency on a real SoC hardware target by up to 7.4x and 3x, respectively with no accuracy drop compared to a network without dilation. It also yields a rich set of Pareto-optimal TCNs starting from a single model, outperforming hand-designed solutions in both size and accuracy.
@article{arxiv.2203.14768,
title = {Pruning In Time (PIT): A Lightweight Network Architecture Optimizer for Temporal Convolutional Networks},
author = {Matteo Risso and Alessio Burrello and Daniele Jahier Pagliari and Francesco Conti and Lorenzo Lamberti and Enrico Macii and Luca Benini and Massimo Poncino},
journal= {arXiv preprint arXiv:2203.14768},
year = {2022}
}