Signed Binary Weight Networks
This addresses the problem of deploying AI with low power and latency requirements, though it appears incremental as it builds on existing sparsity and binarization techniques.
The paper tackles the problem of efficient deep neural network inference by proposing signed-binary networks, which combine weight sparsity and binarization to reduce power and latency while maintaining accuracy. The method achieves comparable accuracy on ImageNet and CIFAR-10 with binary weights and 69% sparsity, leading to real speedups on general-purpose devices and further energy reductions on ASICs.
Efficient inference of Deep Neural Networks (DNNs) is essential to making AI ubiquitous. Two important algorithmic techniques have shown promise for enabling efficient inference - sparsity and binarization. These techniques translate into weight sparsity and weight repetition at the hardware-software level enabling the deployment of DNNs with critically low power and latency requirements. We propose a new method called signed-binary networks to improve efficiency further (by exploiting both weight sparsity and weight repetition together) while maintaining similar accuracy. Our method achieves comparable accuracy on ImageNet and CIFAR10 datasets with binary and can lead to 69% sparsity. We observe real speedup when deploying these models on general-purpose devices and show that this high percentage of unstructured sparsity can lead to a further reduction in energy consumption on ASICs.