Image classification method based on Bayesian neural network random addition decomposition structure

Abstract

The invention discloses an image classification method based on a Bayesian neural network random addition decomposition structure, which comprises the following steps: scaling attribute parameters of a Bayesian neural network to obtain scaled attribute parameters, and converting the scaled attribute parameters into random bit stream data; determining the input number of a reference multiplexer based on the network structure of the Bayesian neural network, determining the input number and quantity of a middle multiplexer according to the input number of the reference multiplexer, and determining an inner product scaling factor of an inner product operation unit based on a parameter scaling factor and the input number and quantity of the middle multiplexer; acquiring random bit stream data, determining an inner product operation output result of an inner product operation unit according to the random bit stream data and an intermediate multiplexer, determining an output scaling factor based on an inner product scaling factor and a parameter scaling factor, and then outputting a final output result of the Bayesian neural network so as to complete image classification. Image classification is carried out through adoption of the above classification method, the hardware implementation cost during classification is greatly reduced.

Description

technical field [0001] The invention belongs to the technical field of image classification, in particular to an image classification method based on Bayesian neural network random addition decomposition structure. Background technique [0002] Image classification is currently the hottest research direction in the field of artificial intelligence and computer vision. In recent years, the use of neural networks to deal with image classification problems has become the mainstream. [0003] In the prior art, the Bayesian neural network is generally used for image classification, and the Bayesian neural network is used for image classification in the prior art, which needs to be combined with a Gaussian random number generator, a multiplier and an adder to work together. Hardware implementation cost prohibitive when sorting. [0004] Therefore, how to reduce the hardware implementation cost of image classification and improve user experience when using Bayesian neural network ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the technical problem that the cost of hardware implementation is too high when using Bayesian neural network for image classification in the prior art, and proposes a method for image classification based on Bayesian neural network random addition decomposition structure

Images