SYSTEM AND METHOD FOR A QUANTIZED NEURAL NETWORK

• Main Authors: KISILEV PAVEL, CHOUKROUN YONI, ZIBULEVSKY MICHAEL
• Format: Patent
• Language: Chinese; English
• Published: 08.06.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

A system for operating a neural network, comprising a processing unit adapted to: receive input data; input the input data to a first layer of a plurality of layers of a neural network; in each of a plurality of iterations compute an output of a layer by: partitioning an activation matrix of the layer, received from a previous layer, into a plurality of sub-matrices; computing a uniform quantization of each of the plurality of sub-matrices to produce a plurality of quantized sub- matrices; combining the plurality of quantized sub-matrices to produce a quantized activation matrix; and computing a matrix product of a plurality of weight values of the layer and the quantized activation matrix to produce the output of the layer; and predict an output value in response to the input data and/or classify a finding detected in the input data, according to an output of a last layer. 一种用于操作神经网络的系统，包括处理单元，所述处理单元用于：接收输入数据；将所述输入数据输入到神经网络的多个层中的第一层；在多个迭代中的每一个中，通过以下操作计算多个层中的层的输出：将从前一层接收到的所述层的激活矩阵划分成多个子矩阵；计算所述多个子矩阵中的每一个的统一量化，