METHOD AND DEVICE FOR ON-DEVICE CONTINUAL LEARNING OF NEURAL NETWORK WHICH ANALYZES INPUT DATA BY OPTIMIZED SAMPLING OF TRAINING IMAGES AND METHOD AND DEVICE FOR TESTING THE NEURAL NETWORK FOR SMARTPHONES DRONES VESSELS OR MILITARY PURPOSE

摘要

The present invention is a method for an on-device continuous learning (Neural Network) of analyzing the input data, the present invention is provided for a smart phone, drone, ship or military purpose, a learning device A, (a) uniformly sampling new data to have a first volume, and causing the boosting network to use a k-dimensional random vector, a previous data corresponding to the k-dimensional correction vector and used for training (Previous Data) Corresponding to, to repeat the process of outputting the data before the synthesis of the second volume, generating a batch (Batch) used in the current learning (Current-Learning); And (b) causing the neural network to generate output information corresponding to the first batch. The present invention is characterized by preventing privacy breaches, optimizing resources such as storage, and training image sampling. It can be performed for process optimization, and can be performed through a learning process of GAN (Generative Adversarial Network).