Device and method for artificial nerve network computing

Abstract

The invention provides a device and a method for artificial nerve network computing. The device for artificial nerve network computing comprises a mapping unit which receives input nerve cell and weight, generates connecting data of the input nerve cell and output nerve cell, and outputs the input nerve cell and the weight after mapping. The corresponding relation between the input nerve cell andthe weight after mapping is an input nerve cell-weight pair. The mapping unit comprises a first mapping unit, used for removing the weight which is 0 or less than a first threshold value, and a secondmapping unit, used for removing the input nerve cell which is 0 or less than a second threshold value.

Description

technical field [0001] The present invention relates to the technical field of data processing, and more particularly relates to an artificial neural network operation device and method. 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 data between a large number of internal nodes. The algorithm used in neural networks is vector multiplication, and symbolic functions and their various approximations are widely used. [0003] Neural networks are widely used in various application scenarios: computational vision, speech recognition, and natural language processing, etc. In recent years, the si...

AI Technical Summary

Problems solved by technology

However, since the GPU is a device specially used to perform graphics and image calculations and scientific calculations, there is no special support for artificial neural network operations, and a large amount of front-end decoding work is still required to perform artificial neural network operations, which brings a lot of additional overhead.

In addition, the GPU has only a small on-chip cache, and the model data (weights) of the multi-layer artificial neural network need to be repeatedly moved from off-chip. The off-chip bandwidth has become the main performance bottleneck, and it has brought huge power consumption overhead.

Images