Method and device for determining neural network models for multiple environments

The invention discloses a method and a device for determining a neural network model for multiple environments. The method comprises the following steps: inputting local environment data of a Ki < th > client into (N-1) first neural network models respectively to obtain (N-1) groups of first training estimated values; determining a first loss value corresponding to the (Ki) th neural network model according to the (N-1) th group of first training estimated values and the (Ki) th group of training estimated values; and under the condition that the first loss value corresponding to the Ki-th neural network model meets a preset convergence condition, determining the Ki-th neural network model as a target neural network model. The technical problem that data in different clients cannot be processed by using the same neural network model under the condition that local data of different clients are not exposed is solved. 本发明公开了一种用于多个环境的神经网络模型的确定方法及装置。其中，该方法包括：将第Ki个客户端本地的环境数据分别输入到N-1个第一神经网络模型，得到N-1组第一训练预估值；根据N-1组第一训