Data Processing and Scheduling Strategies Based on Deep Neural Networks in Cloud-Edge Collaborative Environments

Cloud-edge collaborative environment-based deep neural network data processing and task scheduling is one of the key technologies to realize real-time response and efficient resource utilization. In this study, a multi-level and adaptive deep neural network architecture is constructed based on the deep integration and innovation of the existing hierarchical neural network and the multi-objective optimization algorithms: firstly, the improved convolutional network is used to extract high-dimensional features at the edge, and the recurrent network with variable sequence is introduced to capture the time series association of task data. Secondly, the transmitted features are adaptively processed in the cloud using attention impairment and multi-type data fusion modules, so as to enhance the data communication and the cross-node preparation. At last, through integrating the dynamic resource scheduling strategy with distributed training optimization technology, the low-latency processing and refined management of large-scale and heterogeneous data are achieved. This study has improved data fusion mode, network structure depth and resource scheduling strategy compared with traditional cloud-edge collaboration model, which greatly improve the task allocation and information processing ability of the model in the complex and changeable cloud-edge environment. Experimental results indicate that the proposed model has achieved significant improvement compared with existing mainstream schemes on task delay, through resource utilization and data processing accuracy.