Neural network processing method based on principal component analysis for dimensionality reduction and/or dimensionality enhancement

Abstract

The invention provides a principal component analysis-based dimension-reduced and / or dimension-raised neural network processing method. The principal component analysis-based neural network processingmethod comprises the steps of carrying out dimension reduction treatment on under-chip data and sending the data to an on-chip part; carrying out dimension raising treatment on data after being subjected to dimension reduction treatment and sent to the on-chip part; receiving data obtained after the dimension raising treatment, and executing the neural network operation; carrying out dimension reduction treatment on data obtained after the neural network operation, and sending obtained data to an under-chip part; carrying out dimension raising treatment on data after being subjected to dimension reduction treatment and sent to the under-chip part, and storing obtained data as under-chip data. The principal component analysis is used for carrying out dimension reduction treatment / dimension raising treatment on data. According to the disclosure of the invention, by adopting the principal component analysis-based dimension-reduced and / or dimension-raised neural network processing method, the data dimension reduction treatment is carried out during the data loading and storing process. Therefore, the number of IOs is reduced. The time and energy expenditure is lowered.

Description

technical field [0001] The disclosure belongs to the field of computer technology, and more specifically relates to a neural network processing system and method based on principal component analysis. Background technique [0002] Artificial Neural Networks (ANNs), referred to as Neural Networks (NNs) for short, is an algorithmic mathematical model that imitates the behavioral characteristics of animal neural networks and performs distributed parallel information processing. This kind of network depends on the complexity of the system, and achieves the purpose of processing information by adjusting the interconnection relationship between a large number of internal nodes. The concept of deep learning originated from the research of artificial neural networks. A multi-layer perceptron with multiple hidden layers is a deep learning structure. Deep learning combines low-level features to form more abstract high-level representation attribute categories or features to discover...

AI Technical Summary

Problems solved by technology

[0003] Various current neural network computing devices often face the problem of memory access bottlenecks, and loading and storing data causes a lot of time and energy overhead

Images