MixPath: A Unified Approach for One-shot Neural Architecture Search
This work addresses a bottleneck in neural architecture search for multi-path structures, offering a novel method that improves efficiency and accuracy in model design.
The paper tackles the problem of efficiently searching for multi-path neural architectures by proposing MixPath, a one-shot approach that uses Shadow Batch Normalization to stabilize training and improve ranking accuracy, resulting in state-of-the-art performance on ImageNet.
Blending multiple convolutional kernels is proved advantageous in neural architecture design. However, current two-stage neural architecture search methods are mainly limited to single-path search spaces. How to efficiently search models of multi-path structures remains a difficult problem. In this paper, we are motivated to train a one-shot multi-path supernet to accurately evaluate the candidate architectures. Specifically, we discover that in the studied search spaces, feature vectors summed from multiple paths are nearly multiples of those from a single path. Such disparity perturbs the supernet training and its ranking ability. Therefore, we propose a novel mechanism called Shadow Batch Normalization (SBN) to regularize the disparate feature statistics. Extensive experiments prove that SBNs are capable of stabilizing the optimization and improving ranking performance. We call our unified multi-path one-shot approach as MixPath, which generates a series of models that achieve state-of-the-art results on ImageNet.