Continuous Particle Swarm Optimization-Based Deep Learning Architecture Search for Hyperspectral Image Classification

Deep convolutional neural networks (CNNs) are widely used in hyperspectral image (HSI) classification. However, the most successful CNN architectures handcrafted, which need professional knowledge and consume a very significant amount of time. To automatically design cell-based for HSI classification, we propose an efficient continuous evolutionary method, named CPSO-Net, can dramatically accelerate optimal architecture generation by optimization weight-sharing parameters. First, SuperNet with all candidate operations is maintained to share parameters individuals optimized collecting gradients population. Second, novel direct encoding strategy devised encode into particles, inherit from SuperNet. Then, particle swarm search deep swarm. Furthermore, experiments limited training samples based on four biased unbiased datasets showed that our proposed method achieves good performance comparable state-of-the-art classification methods.