Recognition and Optimization Algorithm of MNIST Dataset Based on LeNet5 Network Structure

摘要

This paper analyzes the content composition and detailed structure of the MNIST dataset, introduces the overall level of the traditional LeNet5 network, and proposes an improved algorithm for its hidden layer, optimization function, activation function, etc. of the network layer structure. Optimization, and the effectiveness of the algorithm is verified by experiments. This paper has certain reference value for traditional neural network optimization and improving the correct classification rate of MNIST data sets.