Spatial Correlation and Value Prediction in Convolutional Neural Networks
This work addresses the compute-intensive nature of CNNs, which is a bottleneck for deployment in resource-constrained environments, but it is incremental as it builds on existing methods for reducing operations.
The paper tackled the problem of high computational cost in convolutional neural networks (CNNs) by proposing a value prediction method that exploits spatial correlation of zero-valued activations, resulting in a 30.4% reduction in multiply-accumulate operations with minimal accuracy degradation (e.g., 1.7% top-1 accuracy drop on ImageNet).
Convolutional neural networks (CNNs) are a widely used form of deep neural networks, introducing state-of-the-art results for different problems such as image classification, computer vision tasks, and speech recognition. However, CNNs are compute intensive, requiring billions of multiply-accumulate (MAC) operations per input. To reduce the number of MACs in CNNs, we propose a value prediction method that exploits the spatial correlation of zero-valued activations within the CNN output feature maps, thereby saving convolution operations. Our method reduces the number of MAC operations by 30.4%, averaged on three modern CNNs for ImageNet, with top-1 accuracy degradation of 1.7%, and top-5 accuracy degradation of 1.1%.