The invention discloses a contract
recursion method for distributed
tensor data updating, and relates to the technical field of
computer data processing, and the method comprises the following steps: initializing a dynamic adaptive contract cluster, sensing and aggregating the dynamic state of the cluster in real time, executing a contract
recursion update cycle, carrying out
fault tolerance and state restoration in a
recursion process, and carrying out iteration termination judgment. According to the invention, through combination of the dynamic adaptive contract and real-time
cluster state perception, intelligent recursion of distributed
tensor update is realized, the efficiency and stability of large-scale training are significantly improved, and a communication strategy can be dynamically optimized according to a network and computing power state;
state consistency in a complex environment is ensured, conflicts are accurately resolved through version vectors and intelligent arbitration, and elegant
fault tolerance is realized by utilizing prediction compensation; and meanwhile, the method has high adaptability, parameters can be optimized on line to reduce manual tuning cost, and robust and efficient support is provided for distributed
machine learning.