为了自动学习CNN的深度网络结构，网络结构的数量随着网络中间层数量的增加呈指数增长，这启发我们使用遗传算法有效地遍历这个大的搜索空间。我们首先提出一种编码方法，将每个网络结构表示为一个固定长度的二进制字符串，然后通过生成一组随机个体来初始化遗传算法。在每一代中，我们定义标准的遗传操作(如选择、突变和交叉)来消除弱势个体并产生更具竞争力的个体。(In order to automatically learn the deep network structure of CNN, the number of network structure increases exponentially with the increase of the number of network middle layer, which inspires us to use genetic algorithm to effectively traverse this large search space.We first propose a coding method to represent each network structure as a binary string of fixed length, and then initialize the genetic algorithm by generating a set of random individuals.In each generation, we define standard genetic operations (such as selection, mutation, and crossover) to eliminate vulnerable individuals and produce more competitive ones.)