Model compression method, device and equipment and storage medium

• Main Authors: WU BAOYUAN, YANG YUJIU, LIU WEI, FAN YANBO, ZHANG YONG, LI TUANHUI
• Format: Patent
• Language: Chinese; English
• Published: 23.08.2019
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The embodiment of the invention discloses a model compression method and device, equipment and a storage medium, and the method comprises the steps: obtaining a to-be-compressed first model which comprises a convolutional layer; performing point multiplication by using scalar and a convolutional filter in the convolutional layer to generate a decomposed convolutional filter, and generating a second model according to the decomposed convolutional filter; based on an objective function with a cardinal constraint and a binary constraint, training the second model by using the training set to obtain a second model meeting a convergence condition; and determining a compressed model corresponding to the first model according to the second model meeting the convergence condition. According to themethod, filter selection and filter learning are combined into a whole by means of a decomposition convolution filter, the importance of all filters is autonomously measured by means of introduced scalar autonomous learning,