Hardware acceleration implementation architecture for forward prediction of convolutional neural network based on FPGA

Abstract

The invention discloses a hardware acceleration implementation architecture for forward prediction of a convolutional neural network based on an FPGA. Aiming at a specific simplified and optimized convolutional neural network Hcnn, hardware architecture research and implementation for a forward prediction process of the convolutional neural network Hcnn are performed. According to the architecture, main operation units of the convolutional neural network are realized based on an optimized pulsation array. Operation processing time and resource consumption are comprehensively considered, and using methods of parallel-serial conversion, data fragmentation, pipeline design and the like, the forward prediction process of the Hcnn convolutional neural network in a parallel pipeline form is realized by taking the degree of parallelism as high as possible and the resource consumption as low as possible as the principle. And the characteristics of data parallelism and assembly line parallelism of the FPGA are fully utilized. The pulsation array structure balances IO read-write and calculation, the throughput rate is improved under the condition that less storage bandwidth is consumed, and the problem that the data access and storage speed is much higher than the data processing speed of a convolutional neural network FPGA is effectively solved.

Description

technical field [0001] The present invention relates to the field of deep learning, one of the important development directions in artificial intelligence, and specifically relates to a hardware acceleration implementation architecture of FPGA-based convolutional neural network forward prediction. Background technique [0002] In recent years, the field of artificial intelligence, especially machine learning, has achieved breakthrough achievements in both theory and application. Deep learning is one of the most important development directions of machine learning. Deep learning can learn features with multi-level abstraction. Therefore, deep learning has excellent performance in solving complex and abstract learning problems. However, as the problem continues to become more complex and abstract, the model of the deep learning network becomes more complex, and the learning time of the model also increases. For example, Google's "AlphaGo" uses a multi-layer neural network str...

AI Technical Summary

Problems solved by technology

Therefore, FPGA is a very good choice for deep learning acceleration. However, there are not many studies on the specific architecture of FPGA implementation of deep learning algorithms. There are problems such as insufficient storage bandwidth, and there is still a lot of room for improvement in acceleration effects.

Images