Neural network parameter adjustment method, device and equipment

Abstract

The embodiment of the invention discloses a neural network parameter adjustment method, device and equipment, and belongs to the technical field of computers. The method comprises the steps of acquiring parameters and a training sample of a trained neural network with the training sample comprising input data, and the parameters and the input data are floating-point numbers with the same precision; for the ith layer of network in the neural network, performing predetermined operation on the ith layer of input data and the ith layer of parameters to obtain a first operation result; Respectivelyconverting the ith layer of input data and the ith layer of parameters into fixed-point numbers conforming to decimal bit widths of the ith layer of network, and performing predetermined operation onthe converted ith layer of input data and ith layer of parameters to obtain a second operation result; And when an error between the second operation result and the first operation result is smallerthan an error threshold, determining the converted ith-layer parameter as a target parameter of the ith-layer network. According to the embodiment of the invention, the efficiency of parameter adjustment can be improved on the premise of ensuring the precision of the neural network.

Description

technical field [0001] The embodiments of the present application relate to the field of computer technology, and in particular to a method, device and equipment for adjusting parameters of a neural network. Background technique [0002] Neural network has high recognition accuracy and good parallelism, and has been widely used in image recognition, object classification, pattern recognition and other fields in recent years. Among them, the application of neural network involves two stages of training and reasoning. Training refers to the use of massive training samples to train the parameters in the neural network; reasoning refers to the use of a neural network with trained parameters to process input data. Get the inference result. [0003] Since the training stage has high requirements for data accuracy and the neural network has a certain robustness, the requirements for data accuracy in the inference stage can be reduced to reduce the hardware resources used in the in...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a parameter adjustment method, device and equipment of a neural network, which are used to solve the p

Images