METHOD AND APPARATUS FOR CORRECTING NEW MODEL ON BASIS OF MULTIPLE SOURCE MODELS, AND COMPUTER DEVICE

摘要

The application relates to the field of artificial intelligence, and provides a method and apparatus for correcting a new model on the basis of multiple source models, and a computer device. The method comprises: inputting first training data into a new model to obtain a first current vector; inputting the first training data into each of a plurality of preset source models, and performing calculation to obtain feature vectors corresponding to the source models; performing fusion calculation to obtain an index vector; calculating a gradient value between the first current vector and the index vector; and adjusting parameters in the new model on the basis of the gradient value. In the present application, training data is inputted into multiple existing source models to obtain multiple feature vectors corresponding thereto, an index vector is then obtained by means of fusion calculation, a gradient value between the index vector and a current vector obtained from a new model are calculated, and parameters in the new model are adjusted on the basis of the gradient value. The new model integrates the advantages of fusing multiple source models, while avoiding the problem of the increased volume and reduced speed that arise from direct use of a fusion model.