CONVOLUTIONAL NEURAL NETWORK MODEL COMPRESSION METHOD, APPARATUS AND DEVICE, AND STORAGE MEDIUM

摘要

Provided are a convolutional neural network model compression method, apparatus and device, and a storage medium. The method comprises: copying an original convolutional neural network model in an application program to obtain N candidate models Mi; compressing any two layers of convolutional kernels of each of the candidate models Mi, and training the candidate models Mi to obtain adjusted candidate models Mi; selecting an optimal candidate model Mk with the least performance loss to run the application program to obtain the current internal environment parameters of a mobile terminal; and taking the optimal candidate model Mk, which meets a preset resource condition, as a compressed convolutional neural network model, otherwise, taking the optimal candidate model Mk as the original convolutional neural network model of the next round of model compression, and re-compressing same. The method also relates to blockchain technology, and the original convolutional neural network model is stored in a blockchain. The method achieves the automatic adaptation of a convolutional neural network model to a mobile terminal for compression.