The invention provides a neural
network model reference neural cognitive mechanism and
machine learning mathematical method, and relates to the field of
machine learning. The method comprises the following steps: S1, establishing a sample
library for sample information, storing the sample
library, generating adversarial samples through a fast gradient symbol method, and dividing the adversarial samples into a
training set and a
test set; S2, enhancing the migration ability from the adversarial sample
library to the
training set through lifelong continuous migration learning; S3, connecting a
recurrent neural network through a
convolutional neural network, and then connecting a BP neural network and a naive Bayes
algorithm to establish a neural
network model; and S4, classifying the target information in the adversarial sample library to form a data table. The adversarial samples are generated through the fast gradient symbol method, videos, images and texts are combined, real scene dialogues, dialogues with images and text expression are more specific, the
artificial intelligence application range is widened, the observation thinking and expression ability is improved, and generalization is improved through different structures.