Uber Deep Learning ETA分享notes

 

The DeepETA team tested and tuned 7 different neural network architectures: MLP [4], NODE [5], TabNet [6], Sparsely Gated Mixture-of-Experts [7], HyperNetworks [8], Transformer [9] and Linear Transformer [10].

We found that an encoder-decoder architecture with self-attention provided the best accuracy.

Encoder with Self-Attention