Device used for executing forward operations of neural network of fully-connected layers and methods

Abstract

The invention provides a device used for executing forward operations of fully-connected layers of an artificial neural network. The device includes an instruction storage unit, a controller unit, a data access unit, an interconnection module, a main operation module and a plurality of slave operation modules. One or more layers of forward operations of the fully-connected layers of the artificial neural network can be realized by using the device. For each layer, weighted summation is firstly carried out on an input neuron vector to calculate an intermediate result vector of the current layer, then bias is added to the intermediate result vector and activation are carried out to obtain an output neuron vector, and the output neuron vector is used as an input neuron vector of a next layer.

Description

technical field [0001] The present invention generally relates to artificial neural networks, and in particular to a device and method for performing forward operation of fully connected layer artificial neural networks. 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. The artificial neural network involves a variety of algorithms, among which the fully connected layer, as an important algorithm in the artificial neural network, is widely used in various artificial neural network models. [0003] One known method to support forward operation of the fully connected layers of a multi-layer artificial neural network is to use a...

AI Technical Summary

Problems solved by technology

Since the GPU is a device specially used to perform graphics and image calculations and scientific calculations, without special support for multi-layer artificial neural network operations, it still requires a lot of front-end decoding work to perform multi-layer artificial neural network operations, which brings a lot of problems. additional cost

In addition, the GPU only has 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, and the off-chip bandwidth has become the main performance bottleneck.

Images