An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

Code

• github.com/locuslab/TCN
  OfficialIn paperpytorch★ 0
• github.com/WenjieDu/PyPOTS
  pytorch★ 1,970
• github.com/proroklab/popgym
  pytorch★ 213
• github.com/Baichenjia/Tensorflow-TCN
  tf★ 129
• github.com/rvandewater/yaib
  pytorch★ 94
• github.com/sindhura97/STraTS
  pytorch★ 81
• github.com/ratschlab/HIRID-ICU-Benchmark
  pytorch★ 60
• github.com/zhong110020/pytorch_TCN
  pytorch★ 24
• github.com/tw-yuhsi/a-new-perspective-for-shuttlecock-hitting-event-detection
  pytorch★ 21
• github.com/wish44165/A-New-Perspective-for-Shuttlecock-Hitting-Event-Detection
  pytorch★ 21

Abstract

For most deep learning practitioners, sequence modeling is synonymous with recurrent networks. Yet recent results indicate that convolutional architectures can outperform recurrent networks on tasks such as audio synthesis and machine translation. Given a new sequence modeling task or dataset, which architecture should one use? We conduct a systematic evaluation of generic convolutional and recurrent architectures for sequence modeling. The models are evaluated across a broad range of standard tasks that are commonly used to benchmark recurrent networks. Our results indicate that a simple convolutional architecture outperforms canonical recurrent networks such as LSTMs across a diverse range of tasks and datasets, while demonstrating longer effective memory. We conclude that the common association between sequence modeling and recurrent networks should be reconsidered, and convolutional networks should be regarded as a natural starting point for sequence modeling tasks. To assist related work, we have made code available at http://github.com/locuslab/TCN .

Tasks

Benchmark Results

Reproductions