Neural network training and face detection method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a neural network training and face detection method and device, equipment and a storage medium. The training method comprises the steps of determining a neural network; according to a first optimization mode, training the neural network at a first learning rate, wherein the first learning rate is updated when the neural network is trained each time; mapping the first learning rate of the first optimization mode into a second learning rate of a second optimization mode in the same vector space; determining that the second learning rate meets a preset updating condition; and continuing to train the neural network at the second learning rate according to a second optimization mode. According to the embodiment, through the mapping of the learning ratein the same vector space, the neural network can be trained by switching the appropriate optimization modes in different stages, the advantages of the appropriate optimization modes can be exerted indifferent stages, the problems generated by other optimization modes are reduced or avoided, and the requirements of two or two aspects for training the neural network are met at the same time.

Description

technical field [0001] Embodiments of the present invention relate to deep learning technologies, and in particular to a neural network training and face detection method, device, device, and storage medium. Background technique [0002] At present, deep learning methods based on neural networks have a wide range of applications in many fields such as computer vision, natural language processing, and text understanding, and these fields basically cover image and video processing, speech processing, and text processing required by current Internet applications. technology. [0003] Deep learning uses the neural network as a data feature extraction tool, trains the parameters in the neural network through a large number of samples, and fits the labels of the samples, such as types, so as to have the predictive ability in scenarios similar to the sample distribution. [0004] In general, users set learning goals, such as labels for classification, location and size of label bo...

AI Technical Summary

Problems solved by technology

[0007] Most of the existing neural network optimization methods use the learning rate and the result of the objective function to update the parameters. The learning rate of the optimization method determines the magnitude of the parameter update. The learning rate has a great influence on the training of the neural network. The existing neural network Usually, a single optimization method is used for training. While meeting certain requirements, it is easy to ignore other requirements

[0008] For example, the learning rate has an impact on the speed and generalization ability of the training neural network: if the learning rate is too small, the training speed of the neural network will be slow, resulting in a long training period and affecting the training efficiency of the neural network; if the learning rate is too large , it is likely to skip the optimal parameters, and the generalization ability of the neural network is poor

Images