基于BP神经网络学习率优化的研究

The Study of Learning Rate Based on BP Neural Network

现今社会人工智能技术快速迭代发展,其应用也越发广泛,包括搜索、数学优化、逻辑推演等工具都应用了人工智能技术。神经网络作为人工智能的重要方法正在被不断地深入研究,而BP神经网络是经典的神经网络之一,在语音分析、图像识别、数字水印、计算机视觉等应用领域都取得了显著的效果。在对BP神经网络进行训练时,学习率的设置是众多参数中至关重要的一项。学习率选取不当将直接导致模型收敛速度慢、模型易越过全局极小值点等问题。针对BP神经网络中的学习率选取开展研究,将传统的固定学习率优化为变化学习率,从而有效地提高了BP神经网络模型的收敛速度以及精确度。

Nowadays, artificial intelligence iterative and development rapidly, many tools was applied artificial intelligence such as Search, mathematical optimization, logic deduction and so on, however, the BP neural network is one of the classic neural network, and it has a remarkable effect in the field of voice analysis, image recognition, digital watermarks, and computer-vi- sion applications. While training BP network how to set learning rate is the most important of many parameters in the course of training. The selection of the learning rate is a direct result of slowing down the model and slowing the model to the point of the global minimum. So, this is a study of the learning rate in the BP neural network, and it＇s a good idea to improve the rate and accuracy of the BP neural net model by optimizing the traditional rate of fixed learning rate to change the learning rate.