Memory Bounded Deep Convolutional Networks
This addresses memory constraints for deploying CNNs in resource-limited environments, but it is incremental as it builds on existing regularization techniques.
The authors tackled the problem of high memory consumption in Convolutional Neural Networks (CNNs) by using sparsity-inducing regularizers during training, which reduced memory usage by a factor of four on AlexNet with minimal accuracy loss.
In this work, we investigate the use of sparsity-inducing regularizers during training of Convolution Neural Networks (CNNs). These regularizers encourage that fewer connections in the convolution and fully connected layers take non-zero values and in effect result in sparse connectivity between hidden units in the deep network. This in turn reduces the memory and runtime cost involved in deploying the learned CNNs. We show that training with such regularization can still be performed using stochastic gradient descent implying that it can be used easily in existing codebases. Experimental evaluation of our approach on MNIST, CIFAR, and ImageNet datasets shows that our regularizers can result in dramatic reductions in memory requirements. For instance, when applied on AlexNet, our method can reduce the memory consumption by a factor of four with minimal loss in accuracy.