A Neural Network Processing System Based on Wavelet Transform to Reduce IO Cost

Abstract

The disclosure provides a neural network processing system for reducing IO overhead based on wavelet transform. The neural network processing system based on wavelet transform comprises: an off-chip compression unit that is used for compressing off-chip date and sending the off-chip date to a chip; and an on-chip computation device that is connected with the off-chip compression unit, and is usedfor receiving the date that is compressed and sent to the on-chip, and executing neural network operation, wherein the compression unit compresses, based on the wavelet transform, the off-chip data. According to the neural network processing system of the disclosure, by compressing the data when the data is loaded and stored, the IO quantity and the time and energy overhead are reduced.

Description

technical field [0001] The present disclosure belongs to the field of computer technology, and more particularly relates to a neural network processing system and method based on wavelet transform. Background technique [0002] Artificial Neural Networks (ANNs) are referred to as Neural Networks (NNs) for short. It is an algorithmic mathematical model that imitates the behavioral characteristics of animal neural networks and performs distributed parallel information processing. This kind of network relies 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 study of artificial neural networks. A multilayer 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] The current various neural network on-chip computing devices often face the problem of memory access bottlenecks, and loading and storing data has caused a lot of time and energy overhead.

Images