Einconv: Exploring Unexplored Tensor Network Decompositions for Convolutional Neural Networks
This work addresses the lack of comprehensive comparisons and optimal decomposition selection for CNN compression, offering potential improvements in efficiency for deep learning applications.
The study systematically explores tensor network decompositions for CNNs, identifying a class that includes known and novel modules, and finds that some nonlinear decompositions outperform existing ones in accuracy and complexity tradeoffs.
Tensor decomposition methods are widely used for model compression and fast inference in convolutional neural networks (CNNs). Although many decompositions are conceivable, only CP decomposition and a few others have been applied in practice, and no extensive comparisons have been made between available methods. Previous studies have not determined how many decompositions are available, nor which of them is optimal. In this study, we first characterize a decomposition class specific to CNNs by adopting a flexible graphical notation. The class includes such well-known CNN modules as depthwise separable convolution layers and bottleneck layers, but also previously unknown modules with nonlinear activations. We also experimentally compare the tradeoff between prediction accuracy and time/space complexity for modules found by enumerating all possible decompositions, or by using a neural architecture search. We find some nonlinear decompositions outperform existing ones.