Federal learning-based model training processing method and device, equipment and medium

The invention relates to the technical field of artificial intelligence, and discloses a federated learning-based model training processing method and device, equipment and a storage medium, which are used for reducing the calculation cost. The method comprises the following steps: sending a local initial network model of central node equipment to each piece of child node equipment; target parameters uploaded by the child node devices are obtained, parameters of a first preset feature parameter layer of a trained network model are obtained after the initial network model is trained for the child node devices, the first preset feature parameter layer is a network layer with the receptive field larger than a preset value in the trained network model, and the first preset feature parameter layer is a network layer with the receptive field larger than the preset value; parameters of a second preset feature parameter layer of the trained network model are frozen locally by the sub-node device, and the second prese