Implement a continual learning framework that can incrementally update classifiers and efficiently relabel old datasets without the cost of reprocessing the existing database. Significance: In previous solution, the classifiers are entirely retrained when the database gets updated by new information – this is redundant and wastes time and computation. Such a framework can greatly reduce the computational redundancies and make instant decisions using latest information.
Enable Semi-Supervised Learning and novel classes detection. Significance: One would like to compare performance of new data to older data and making these comparisons is impossible since the data was classified with older knowledge. By semi-supervised learning, we can make a unified system that will leverage labeled and unlabeled data to incrementally update the classifier and previous datasets. This will enable one to classify on a large volume data, which occurs in healthcare, social networks.