Artificial intelligence-based neural network network parameter processing method and device

Abstract

The invention provides a neural network parameter processing method and device based on artificial intelligence. According to the embodiment, N random seeds adopted by each element identifier in M element identifiers of neural network parameters are obtained, N compression parameters of each element identifier are further obtained according to each element identifier and the N random seed adopted by the element identifier, and a reconstruction neural network is utilized to obtain reconstruction elements corresponding to each element identifier according to the N compression parameters of each element identifier, so that the neural network parameters can be obtained according to the reconstruction elements corresponding to each element identifier; and in the calculation process of the neural network parameters, the plurality of random seeds are used to carry out a plurality of times of random operation, a plurality of compression parameters are obtained, the reconstruction neural network is used again, and the reconstruction elements of the neural network parameters are calculated according to the compression parameters. Only a certain number of compression parameters such as compression space information of eight compression parameters are set, and specific elements of the neural network parameters are not set respectively, and the neural network parameter of the neural network are effectively reduced, so that the memory resource consumption of the neural network is lowered.

Description

[0001] 【Technical Field】 [0002] The present invention relates to Internet technology, in particular to a method and device for processing network parameters of a neural network based on artificial intelligence. [0003] 【Background technique】 [0004] Artificial Intelligence (Artificial Intelligence), the English abbreviation is AI. It is a new technological science that studies and develops theories, methods, techniques and application systems used to simulate, extend and expand human intelligence. Artificial intelligence is a branch of computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, language recognition, image recognition, Natural language processing and expert systems, etc. [0005] With the increase of training data for neural networks, especially deep neural networks, in order to better learn the characteristics of the...

AI Technical Summary

Problems solved by technology

[0005] With the increase of training data of neural networks, especially deep neural networks, in order to better learn the effect of data feature enhancement, the neural network parameters used to represent the neural network increase rapidly, which will lead to a rapid increase in the consumption of memory resources. This restricts the application scenarios of neural networks, such as mobile phones and other terminals with limited memory resources.

Images