An apparatus and method for executing self-learning operation of artificial neural network

Abstract

The invention provides a device and method for executing the self-learning operation of an artificial neural network. The device comprises a controller unit, an interconnection module, a master operation module and a plurality of slave operation modules. According to the method, the self-learning pre-training of the multi-layer neural network can be carried out according to a layer-by-layer training mode, and for each layer of network, the self-learning pre-training of the layer of network is completed through multiple times of operation iteration until the weight updating is smaller than a certain threshold value. Each iteration process can be divided into four stages, a first-order hidden layer intermediate value, a first-order visible layer intermediate value and a second-order hidden layer intermediate value are respectively calculated and generated at the first three stages, and the weights are updated by utilizing the intermediate values at the first three stages in the last stage.

Description

technical field [0001] The present disclosure relates to artificial neural network technology, in particular to a device and method for performing artificial neural network self-learning operations. Background technique [0002] Multi-layer artificial neural networks are widely used in the fields of pattern recognition, image processing, function approximation, and optimization calculations. In recent years, multi-layer artificial networks have been favored by academic circles and researchers due to their high recognition accuracy and good parallelism. industry is getting more and more attention. [0003] A typical multi-layer artificial neural network training method is the backpropagation (BP) algorithm. This method is a representative type of supervised learning, which requires a large number of labeled training samples during the training process, but the cost of sample collection is very high. At the same time, during the training process of this method, the error cor...

AI Technical Summary

Problems solved by technology

[0006] What the present disclosure aims to solve is that in the prior art, general-purpose processors (GPU, CPU) need a series of simple operations and memory access operations for multi-layer neural network pre-training, the power consumption of front-end decoding is relatively large, and the existing general-purpose processing However, there are many problems such as high data access overhead and low computing performance of a single general-purpose processor.

Images