Method and system for deep learning based on GPU

Abstract

The invention discloses a method and a system for deep learning based on a GPU. The system is a one-of-a-kind system and comprises a CPU and at least a GPU. The method comprises: the CPU transmitting to-be-trained data to each GPU; each GPU using the to-be-trained data to perform forward-backward calculation to obtain weight information of a neural network model, and feeding the weight information back to the CPU; the CPU upgrading the neural network model according to the weight information, and transmitting an upgraded weight neural network model to each GPU, executing the above steps in a looping manner, until a deep learning process of the neural network model is completed. In the above scheme, the GPUs which has strong parallel computing capacity execute time-consuming forward-backward calculation, and a synergetic deploying method of the CPU and a plurality of GPU cards is used, so that problems in the prior art that calculation consumes long time and is low in efficiency, system deployment is complex, and cost is high are solved.

Description

technical field [0001] The invention relates to the fields of high-performance computing, deep learning technology and the Internet, in particular to a GPU-based deep learning method and system. Background technique [0002] Today, deep learning is a new field in machine learning research. Its motivation is to establish and simulate the neural network of human brain for analysis and learning. It imitates the mechanism of human brain to interpret data, such as images, sounds and texts. [0003] In 2006, Geoffrey Hinton, a professor at the University of Toronto in Canada and a leader in the field of machine learning, and his students published an article in the top academic journal "Science", which opened the wave of deep learning in academia and industry. Since 2006, deep learning has continued to gain momentum in academia. Stanford University, New York University, University of Montreal, Canada, etc. have become important centers for deep learning research. In 2010, the DA...

AI Technical Summary

Problems solved by technology

[0006] In view of this, the present invention provides a GPU-based deep learning method and system to solve the problems in the prior art that the calculation time is long, the efficiency is low, the system deployment is complicated, and the cost is high.

Images