Pruning In Time (PIT): A Lightweight Network Architecture Optimizer for Temporal Convolutional Networks

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 pr...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:2021 58th ACM/IEEE Design Automation Conference (DAC) s. 1015 - 1020
Hlavní autori: Risso, Matteo, Burrello, Alessio, Pagliari, Daniele Jahier, Conti, Francesco, Lamberti, Lorenzo, Macii, Enrico, Benini, Luca, Poncino, Massimo
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: IEEE 05.12.2021
Predmet:
Computer architecture
Deep learning
Design automation
Edge Computing
Network architecture
Neural Architecture Search
Neural networks
Temporal Convolutional Networks
Training
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí: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 7.4× and 3×, 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.
DOI:10.1109/DAC18074.2021.9586187