17 сент 2019 г. - Low-rank Embedding of Kernels in Convolutional Neural Networks under Random Shuffling (ICASSP 2019)
Описание:
提出了 randomly-shuffled tensor decompo- sition (RsTD) based convolutional layer不同于传统的 TD-based compression
shows that the kernel can be embedded into more general or even random low-rank subspaces.
结果：
基于 CIFAR-10 分类任务，RsTD 方法在压缩率jiao低(<0.01) 时，相对传统 TD 方法 ACC 更高。
Добавлено на ленту времени:
Дата:
17 сент 2019 г.
Сейчас
~ 6 г назад