Method for extracting data features and related device

Abstract

The invention provides a method and a device for extracting data features in the field of artificial intelligence. In the technical scheme of the invention, a target feature in quantized data is extracted based on a quantized feature extraction parameter by adopting the additive convolution operation, i.e., the sum of absolute values of the difference values of the quantized feature extraction parameter and the quantized data is calculated, so that the target feature is obtained according to the sum; in this way, storage resources can be saved, meanwhile, computing resources can be saved, andtherefore application of artificial intelligence to resource-limited equipment can be reduced. Besides, the invention further provides a method for quantizing feature extraction parameters and the data by using the same quantization parameters, so that the feature data can be inversely quantized on the basis of the quantization parameters when the extracted feature data is inversely quantized, thecalculation complexity can be reduced while the accuracy of the feature data is improved, and calculation resources are saved. Furthermore, the invention also provides a hardware device for realizingthe addition convolution operation.

Description

technical field [0001] This application relates to data computing technology in the field of artificial intelligence, and more specifically, to a method and related devices for extracting data features. Background technique [0002] Artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a branch of computer science that attempts to understand the nature of intelligence and produce a new kind of intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning and decision-making. Research in the field of...

AI Technical Summary

Problems solved by technology

[0007] However, after research, it is also found that when the input information is multiplicatively convolved based on the weight parameters of the quantized neural network, a lot of hardware circuit resources are required, the calculation complexity is high, and the energy consumption is still very large.

That is to say, the large amount of computing resources required for multiplicative convolution of the input information based on the weight parameters of the quantized neural network limits the application of artificial intelligence on devices with limited computing resources.

Images