by Darren Bockman

by Xiongjie Yu

Neural Network on a Tensor Train

Redundancy in the weight matrix of a neural network can be much reduced with a tensor network formulation. More than 99% compression rate can be achieved while maintaing accuracy. Tensor train representation of a neural network is compact. This formalism may allow neural networks to be trained on mobile devices.

Presentation Summary

In this presentation, I present:

  • Basic idea of tensor train representation of a matrix
  • Compression of a feed-forward neural network

Examples

References

All optimization.

ALGORITHM
machine learning