An Algorithm for Automated Search Model with Pruning Structure based on Big Data

A compressed pruning scheme for convolutional neural networks based on automated search is proposed to address the problem that convolutional neural networks consume many computational resources under large data. First, the target constraint is added by hyperparameters to reduce the combination. Then the search process is viewed as an optimization problem, and the pruning structure is searched and optimized automatically by using the sparrow search algorithm. The algorithm improves model compression efficiency by determining the number of channels in each layer of the neural network and avoids rule-based selection. Experimental datas show that the parameter compression rate reaches 55.28% and the FLOPs are reduced by 58.57% on the ResNet-56 dataset; the parameter compression rate reaches 80.61% and the FLOPs are reduced by 70.34% on the VGG-16 data.