Sample incremental learning method based on deep neural network

Abstract

The invention discloses a sample incremental learning method based on a deep neural network, and relates to the technical field of big data intelligent analysis. In this way, the prediction result pj (xk) of the old data xk in the new model is close to the prediction result pj-1 (xk) of the old data xk in the old model as much as possible, the model keeps the recognition capability of the old data while learning the new data, and the model can keep the prediction result of the old data, so that the forgetting of old knowledge is relieved, and the model is prevented from deviating to predict the new data.

Description

technical field [0001] The invention relates to the technical field of big data intelligent analysis, in particular to a sample incremental learning method based on a deep neural network. Background technique [0002] There is no strict definition of sample incremental learning, but its main features include two points: (1) It can add new samples of known classes to the existing knowledge system. (2) It can gradually evolve a basic knowledge system into a more complex system. [0003] For sample incremental learning, previous work mainly adopts various non-neural network learning algorithms, such as support vector machines, decision trees, Bayesian networks, etc. With the development of deep neural networks, the use of deep neural networks to achieve incremental learning similar to human memory has become a hot topic. However, the current research mainly focuses on the incremental learning of categories, while the research on incremental learning of samples is rare. And fe...

Images