Parallel Optimization of Depth Learning Algorithm Based on Behavior Recognition

In order to solve the problem of low parallel processing efficiency of current deep learning algorithm data, a parallel optimization method of deep learning algorithm based on behavior recognition is proposed. The behavior recognition method is used to extract learning morphological features, and the clustering specification parameters of deep learning features are calculated. The learning state feature values and clustering specification parameters are compared according to the calculation results, so as to effectively check out the heterogeneous data hidden in the parallel processing process. The parallel processing execution steps are optimized according to the parameters of the depth learning feature clustering specification, and the control flow of the parallel processing of the behavior depth learning is improved by planning the parallel rules of the depth learning algorithm, so as to finally achieve the research goal of the parallel optimization of the depth learning algorithm based on the behavior recognition. Finally, experiments show that the parallel optimization efficiency of the depth learning algorithm based on behavior recognition is significantly higher than that of the traditional depth learning algorithm.