ESOA Algorithm Based on learning rate optimization in Convolutional neural networks

In order to solve the problem of imperfect learning rate optimization in CNN model, this research proposes an adaptive learning rate update model based on Egret Swarm Optimization Algorithm. The model is mainly composed of three components: : a sit-and-wait strategy, aggressive strategy as well as discriminant conditions. These three components play a pivotal role in the learning rate optimization process. This algorithm improves the operation speed of the whole model, and has a significant improvement in convergence speed and accuracy. The proposed ESOA-CNN are compared to the conventional learning rate annealing algorithm and the Particle Swarm Optimization (PSO). These algorithms were evaluated on learning style classification data in terms of accuracy, precision, etc. The result of experimental further verify the effectiveness of the ESOA-CNN model.