Neural network processor based on mode frequency statistical encoding and design method

Abstract

The invention provides a neural network processor based on mode frequency statistical encoding and a design method, and relates to the technical field of hardware acceleration of neural network model calculation. The processor includes at least one storage unit for storing an operating command and operational data; at least one calculating unit for neural network calculation; a control unit, connected with the at least one storage unit and the at least one calculating unit and used for obtaining the operating command stored in the at least one storage unit through the at least one storage unit and parsing the operating command to control the at least one calculating unit; at least one data compression unit, each data compression unit being connected with the at least one calculating unit and used for compressing calculated results obtained according to the operational data and performing re-encoding based on mode frequency statistics; and at least one data decompression unit, each data decompression unit being connected with the at least one calculating unit and used for decompressing the compressed operational data.

Description

technical field [0001] The invention relates to the technical field of hardware acceleration for neural network model calculation, in particular to a neural network processor and a design method based on pattern frequency statistical coding. Background technique [0002] Deep learning technology has developed rapidly in recent years. Deep neural networks, especially convolutional neural networks, have made great achievements in image recognition, speech recognition, natural language understanding, weather prediction, gene expression, content recommendation and intelligent robots. Wide range of applications. [0003] The deep network structure obtained by deep learning is an operational model, which contains a large number of data nodes, each data node is connected to other data nodes, and the connection relationship between each node is represented by weight. With the continuous improvement of the complexity of neural networks, neural network technology has many problems in...

AI Technical Summary

Problems solved by technology

[0004] In the actual application of the current deep neural network, the network scale is getting larger and larger, the data throughput is getting higher and higher, and the task types are becoming more and more complex, which will lead to larger neural network processor circuit scale, lower data transmission efficiency, and lower computing speed. become worse

In the actual application of the existing technology, there are a large number of data elements with a value of 0 in the neural network calculation process. After data operations such as multiplication and addition, such elements have no numerical impact on the calculation results, but the neural network processor is processing These data elements will occupy a large amount of on-chip storage space, consume redundant transmission resources and increase runtime, so it is difficult to meet the performance requirements of neural network processors

Images