Stable Low-rank Tensor Decomposition for Compression of Convolutional Neural Network
This addresses the issue of numerical instability in compressing overparameterized CNNs for efficient deployment, offering an incremental improvement in stabilization.
The paper tackles the problem of degeneracy in low-rank tensor decomposition for compressing convolutional neural networks, presenting a novel stabilization method that results in much lower accuracy degradation and consistent performance on image classification tasks.
Most state of the art deep neural networks are overparameterized and exhibit a high computational cost. A straightforward approach to this problem is to replace convolutional kernels with its low-rank tensor approximations, whereas the Canonical Polyadic tensor Decomposition is one of the most suited models. However, fitting the convolutional tensors by numerical optimization algorithms often encounters diverging components, i.e., extremely large rank-one tensors but canceling each other. Such degeneracy often causes the non-interpretable result and numerical instability for the neural network fine-tuning. This paper is the first study on degeneracy in the tensor decomposition of convolutional kernels. We present a novel method, which can stabilize the low-rank approximation of convolutional kernels and ensure efficient compression while preserving the high-quality performance of the neural networks. We evaluate our approach on popular CNN architectures for image classification and show that our method results in much lower accuracy degradation and provides consistent performance.