Learning Deeply Shared Filter Bases for Efficient ConvNets

Woochul Kang, Daeyeon Kim

Keywords: Deep learning, ConvNets, parameter sharing, model compression, convolutional neural networks, recursive networks

Abstract: Recently, inspired by repetitive block structure of modern ConvNets, such as ResNets, parameter-sharing among repetitive convolution layers has been proposed to reduce the size of parameters. However, naive sharing of convolution filters poses many challenges such as overfitting and vanishing/exploding gradients, resulting in worse performance than non-shared counterpart models. Furthermore, sharing parameters often increases computational complexity due to additional operations for re-parameterization. In this work, we propose an efficient parameter-sharing structure and an effective training mechanism of deeply shared parameters. In the proposed ConvNet architecture, convolution layers are decomposed into a filter basis, that can be shared recursively, and layer-specific parts. We conjecture that a shared filter basis combined with a small amount of layer-specific parameters can retain, or further enhance, the representation power of individual layers, if a proper training method is applied. We show both theoretically and empirically that potential vanishing/exploding gradients problems can be mitigated by enforcing orthogonality to the shared filter bases. Experimental results demonstrate that our scheme effectively reduces redundancy by saving up to 63.8% of parameters while consistently outperforming non-shared counterpart networks even when a filter basis is deeply shared by up to 10 repetitive convolution layers.

One-sentence Summary: We propose an efficient recursive parameter-sharing structure and an effective training mechanism for ConvNets.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=uy3L-t4ps