LeanConvNets: Low-cost Yet Effective Convolutional Neural Networks
This work addresses efficiency issues in CNNs for applications like image classification and semantic segmentation, offering a method to reduce computational costs with minimal accuracy trade-offs, though it is incremental as it builds on existing networks like ResNets.
The paper tackles the high computational cost of convolutional neural networks (CNNs) by introducing LeanConvNets, which sparsify fully-coupled convolution operators to reduce weights, operations, and latency with minimal accuracy loss. Results show that lean versions of ResNet achieve accuracy close to state-of-the-art networks while being computationally less expensive, often outperforming comparable reduced architectures like MobileNets and ShuffleNets on benchmark image classification and semantic segmentation tasks.
Convolutional Neural Networks (CNNs) have become indispensable for solving machine learning tasks in speech recognition, computer vision, and other areas that involve high-dimensional data. A CNN filters the input feature using a network containing spatial convolution operators with compactly supported stencils. In practice, the input data and the hidden features consist of a large number of channels, which in most CNNs are fully coupled by the convolution operators. This coupling leads to immense computational cost in the training and prediction phase. In this paper, we introduce LeanConvNets that are derived by sparsifying fully-coupled operators in existing CNNs. Our goal is to improve the efficiency of CNNs by reducing the number of weights, floating point operations and latency times, with minimal loss of accuracy. Our lean convolution operators involve tuning parameters that controls the trade-off between the network's accuracy and computational costs. These convolutions can be used in a wide range of existing networks, and we exemplify their use in residual networks (ResNets). Using a range of benchmark problems from image classification and semantic segmentation, we demonstrate that the resulting LeanConvNet's accuracy is close to state-of-the-art networks while being computationally less expensive. In our tests, the lean versions of ResNet in most cases outperform comparable reduced architectures such as MobileNets and ShuffleNets.