APPARATUS FOR LEARNING CLASSIFICATION MODEL BASED ON DEEP LEARNING USING ADVERSARIAL GENERATOR AND METHOD USING THE SAME

摘要

Disclosed are a deep learning-based classification model learning apparatus using hostile disruptors and a method using the same. In the classification model learning method according to an embodiment of the present invention, the classification model learning apparatus extracts at least one feature (FEATURE) used for class classification of input data in the classification model, and based on the hostile interference model, the at least The classification model is generated by generating masking data by modulating one feature, inputting the masking data into the classification model, and updating any one of the hostile interference model and the classification model according to the classification result of the classification model. learn