VECTOR QUANTIZATION DECODING HARDWARE UNIT FOR REAL-TIME DYNAMIC DECOMPRESSION FOR PARAMETERS OF NEURAL NETWORKS

摘要

A convolutional neural network processing system (102) comprises:an input layer configured to receive input data (108),a decompressor unit (106) configured to receive encoded kernel data (113) encoded with a vector quantization process and to generate decompressed kernel data based on the encoded kernel data,a convolutional accelerator (104) configured to receive the decompressed kernel data, to receive feature data based on the input data (108), and to perform a convolution operation (104) on the feature data and the decompressed kernel data; anda fully connected layer configured to receive convolved data from the convolutional accelerator (104) and to generate prediction data (111) based on the convolved data.;Embodiments of a corresponding electronic device include an integrated circuit, a reconfigurable stream switch formed in the integrated circuit along with a plurality of convolution accelerators (104) and a decompression unit (106) coupled to the reconfigurable stream switch. The decompression unit (106) decompresses encoded kernel data in real time during operation of convolutional neural network.