FUSION TRAINING METHOD AND APPARATUS FOR NEURAL NETWORK MODEL

摘要

Provided in the embodiments of the description are a fusion training method and apparatus for a neural network model. The model training process of a neural network model comprises several training periods, each training period corresponds to the process of carrying out model training using all sample data in a training sample set, and the neural network model is used for carrying out service prediction on input service data. In the current first training period, when the first training period is not the very first training period, on the basis of the accumulation of prediction data of first sample data by a neural network model obtained when training in a training period prior to the first training period is finished, first target prediction data is obtained, i.e. a training process of a neural network model to be trained is adjusted according to the first target prediction data, and the neural network model to be trained is updated.