Pruning In Time (PIT): A Lightweight Network Architecture Optimizer for Temporal Convolutional Networks
This provides a lightweight optimization method for TCNs in time-series processing, offering practical benefits for hardware deployment, though it is incremental as it builds on existing TCN frameworks.
The paper tackles the problem of optimizing time-dilated convolutions in Temporal Convolutional Networks (TCNs) by proposing an automatic dilation optimizer that treats it as weight pruning on the time-axis, reducing model size by up to 7.4x and inference latency by 3x with no accuracy drop.
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.