BCNet: Searching for Network Width with Bilaterally Coupled Network
This work addresses a specific bottleneck in neural architecture search for more efficient convolutional neural networks, offering an incremental improvement over existing methods.
The paper tackles the problem of unfair channel training in one-shot supernets for network width search, which affects channel pruning under hardware constraints, and introduces BCNet to ensure each channel is fairly trained, achieving state-of-the-art or competitive performance on CIFAR-10 and ImageNet, with a 0.48% accuracy improvement for EfficientNet-B0 on ImageNet at the same FLOPs.
Searching for a more compact network width recently serves as an effective way of channel pruning for the deployment of convolutional neural networks (CNNs) under hardware constraints. To fulfill the searching, a one-shot supernet is usually leveraged to efficiently evaluate the performance \wrt~different network widths. However, current methods mainly follow a \textit{unilaterally augmented} (UA) principle for the evaluation of each width, which induces the training unfairness of channels in supernet. In this paper, we introduce a new supernet called Bilaterally Coupled Network (BCNet) to address this issue. In BCNet, each channel is fairly trained and responsible for the same amount of network widths, thus each network width can be evaluated more accurately. Besides, we leverage a stochastic complementary strategy for training the BCNet, and propose a prior initial population sampling method to boost the performance of the evolutionary search. Extensive experiments on benchmark CIFAR-10 and ImageNet datasets indicate that our method can achieve state-of-the-art or competing performance over other baseline methods. Moreover, our method turns out to further boost the performance of NAS models by refining their network widths. For example, with the same FLOPs budget, our obtained EfficientNet-B0 achieves 77.36\% Top-1 accuracy on ImageNet dataset, surpassing the performance of original setting by 0.48\%.