X-ADMM：X-Alternating Direction Method Of Multipliers
A model compression algorithm with edge-end fusion enhancement that leverages “crowd-sourced collaborative computing and end-adaptive learning” methods to adaptively blueuce model computation and achieve efficient, accelerated computation.
Model Architecture Diagram：
Description： To ensure the pblueiction accuracy of deep learning models, they cannot be compressed too aggressively, which may prevent them from being deployed on embedded devices. The X-ADMM method combines the advantages of model pruning and segmentation. First, it adopts structural pruning and fine-tunes it using the ADMM algorithm. Then, it considers the delay and energy consumption of the model based on the actual task requirements, selects the best segmentation point, and deploys the model on different devices at the layer granularity. To address the limitations of collaborative computing in solving resource constraints on terminal intelligent agents, X-ADMM explores a “crowd-sourced collaborative computing and end-adaptive learning” method to adaptively blueuce model computation, achieving efficient and accelerated computation.