Cramnet: Layer-wise Deep Neural Network Compression with Knowledge Transfer from a Teacher Network
This addresses efficiency issues for deploying neural networks on resource-constrained devices like mobile hardware.
The paper tackles the computational and memory bottlenecks of neural networks, particularly in mobile applications, by developing a compression method that reduces networks to less than 10% of memory and 25% of computational power without accuracy loss or reliance on sparse networks.
Neural Networks accomplish amazing things, but they suffer from computational and memory bottlenecks that restrict their usage. Nowhere can this be better seen than in the mobile space, where specialized hardware is being created just to satisfy the demand for neural networks. Previous studies have shown that neural networks have vastly more connections than they actually need to do their work. This thesis develops a method that can compress networks to less than 10% of memory and less than 25% of computational power, without loss of accuracy, and without creating sparse networks that require special code to run.