Ship-based neural network algorithm processing topology generation and dynamic maintenance method and device

The invention discloses a ship-borne neural network algorithm processing topology generation and dynamic maintenance method and equipment, and belongs to the field of intelligent computing. Comprising the following steps: after a ship-borne heterogeneous intelligent computing cluster is started, virtualizing computing resources of each physical node according to types to obtain a plurality of virtual computing nodes; for all the virtual computing nodes, the virtual computing nodes of the same type and different models are converged into the same primary resource pool, the primary resource pools of the same type are merged into a secondary resource pool, and all the secondary resource pools are merged into an available resource pool; and after a ship-borne neural network processing task arrives, selecting a plurality of virtual computing nodes from the available resource pool to generate a processing topology for executing the task. According to the invention, efficient execution of the neural network algorith