10 Jul 2018 Jahr - Wide Compression: Tensor Ring Nets (CVPR 2018)

Beschreibung:

Three-component decomposition

Tensor Ring Nets (TRN), in
which layers of a deep neural network are compressed using tensor ring factorization.

结果：
TR-Nets approach is able to compress LeNet-5 by 11× without losing accuracy, and can compress the state-of-the-art Wide ResNet by 243× with only 2.3% degradation in Cifar10 image classification

Zugefügt zum Band der Zeit:

Datum: