Method and device for training neural network for image recognition

• Main Authors: OH YOUNG-MIN, LEE WON JO, CHO MIN-KYOUNG
• Format: Patent
• Language: Chinese; English
• Published: 15.04.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a method and equipment for training a neural network for image recognition. The method comprises the following steps: receiving an input image set; executing the pre-trained neural network by taking the input image set as input to obtain an inferred result value, and obtaining a first task accuracy of the pre-trained neural network based on a predetermined result value corresponding to the input image set and the inferred result value; pruning the neural network based on the channel unit by adjusting weights between nodes of a plurality of channels based on a preset learning weight and based on channel-by-channel pruning parameters corresponding to each channel of a plurality of layers of the pre-trained neural network; updating a learning weight based on the first task accuracy and the task accuracy of the trimmed neural network; updating channel-by-channel trimming parameters based on the updated learning weight and the task accuracy of the trimmed neural network; and based on the ch