Compressing Neural Networks with the Hashing Trick
This addresses the memory constraints of mobile devices for deploying deep learning models, representing an incremental improvement in model compression techniques.
The paper tackles the problem of large neural network models being unsuitable for mobile devices with limited memory by introducing HashedNets, a novel architecture that compresses models using a hashing trick to share weights, resulting in substantial reductions in storage requirements while mostly preserving performance on benchmark datasets.
As deep nets are increasingly used in applications suited for mobile devices, a fundamental dilemma becomes apparent: the trend in deep learning is to grow models to absorb ever-increasing data set sizes; however mobile devices are designed with very little memory and cannot store such large models. We present a novel network architecture, HashedNets, that exploits inherent redundancy in neural networks to achieve drastic reductions in model sizes. HashedNets uses a low-cost hash function to randomly group connection weights into hash buckets, and all connections within the same hash bucket share a single parameter value. These parameters are tuned to adjust to the HashedNets weight sharing architecture with standard backprop during training. Our hashing procedure introduces no additional memory overhead, and we demonstrate on several benchmark data sets that HashedNets shrink the storage requirements of neural networks substantially while mostly preserving generalization performance.