Dynamic large graph digest algorithm of adversarial neural network

The invention discloses a dynamic large graph digest algorithm of an adversarial neural network, which relates to the technical field of image processing, solves the problem of quick and accurate retrieval in an online information retrieval system by using a dynamic adversarial neural network model, and obtains depth features of a large graph through multi-layer convolution and pooling operation. A large amount of complex information in the condensed image is subjected to small-scale and high-density image abstract representation; a pre-training model is constructed in a self-supervised learning mode, and when a downstream task needs to be solved, the pre-training model can be used for fine tuning instead of training a brand new model from the beginning every time, so that the model optimization process is accelerated; according to the method, an efficient feasible scheme for online information retrieval is provided for a user, generalization is good, the optimal graph abstract in the dynamic large graph chan