Research On Automatic CNN Architecture Search Method For Hyperspectral Image Classificatio

Deep Convolutional Neural Network(CNN)shows great effectiveness on Hyperspectral Image(HSI)classification.However,the architecture design of CNN greatly relies on abundant expert knowledge and experience,which poses great prohibition to its easy use in real-world applications.To alleviate the issue,this paper proposes an evolving block-based CNN(EB-CNN)to search the optimal architecture based on the genetic algorithm automatically.Specifically,two kinds of basic blocks with totally six different configurations are first designed to construct the search space from a macro perspective.This can not only reduce the dependence on expert knowledge and experience in the detailed operation of CNN,but also greatly reduce the search space,which is conducive to the efficient search of CNN structure by genetic algorithm.Then,a flexible encoding strategy is devised for the genetic algorithm to allow different chromosomes to evolve with different lengths.In this manner,the width of each layer and the depth of the architecture can be simultaneously optimized.Furthermore,a novel swapping mutation operator is proposed for the genetic algorithm to speed up the search efficiency and save computing resources.At last,to further improve the performance of the algorithm,facing the CNN architecture and aiming at the problem of interfering pixels at the edge of HSI input data,an evolving CNN search method EA-CNN based on 3D attention module is proposed.With the above techniques,the proposed algorithm automatically seeks the optimal CNN architecture for HSI classification,leading to its better usability than handcrafted CNNs.At last,extensive experiments conducted on 5 commonly used HSI datasets demonstrate that the proposed EB-CNN and EA-CNN achieves better performance,as compared to state-of-the-art peer algorithms.