Compiler techniques to speed-up deep learning have seen a lot of recent research activity. The article is about Intel's nGraph and provides comparisons to Googl

Compiler techniques to speed-up deep learning have seen a lot of recent research activity. The article is about Intel's nGraph and provides comparisons to Google's XLA: https://www.tensorflow.org/performance/xla/ https://www.tensorflow.org/performance/xla/ Amazon and University of Washington have NNVM: https://aws.amazon.com/blogs/machine-learning/introducing-nnvm-compiler-a-new-open-end-to-end-compiler-for-ai-frameworks/ https://aws.amazon.com/blogs/machine-learning/introducing-nn... And there's University of Illinois's DLVM: https://arxiv.org/abs/1711.03016 https://arxiv.org/abs/1711.03016