A method for training a deep learning network based on AI and a learning device using the same

摘要

The present invention provides a method for training a deep learning network based on artificial intelligence (AI), (a) a learning device inputs unlabeled data to an active learning network to cause the active learning network to use the unlabeled data Among the existing learning networks - the existing learning network is in a learned state using the existing labeled data - to extract sub unlabeled data determined as hard examples useful for learning of the sub unlabeled data inputting the labeled data into an auto-labeling network to cause the auto-labeling network to label each of the sub-unlabeled data to generate new labeled data; (b) the learning device causes the continuous learning network to sample the new labeled data and the existing labeled data to generate a mini-batch including the sampled new labeled data and the sampled existing labeled data and train the existing learning network using the mini-batch, the new sampled labeled data are all reflected in learning, and the sampled existing labeled data is only used when the performance of the existing learning network is low generating a learned learning network by learning the existing learning network by reflecting the learning; and (c) the learning device causes the explanable analysis network to generate insightful results for verification data through the learned learning network, and transmits the insightful results to at least one human engineer. transmits an analysis result that the human engineer analyzes the performance of the learned learning network with reference to the Insight Pearl results, and at least one of the active learning network and the continuous learning network with reference to the analysis result modify and improve; It relates to a method comprising