Applied Deep Learning

Using machine learning in real-world problems often requires more skills than those needed to apply machine learning in academic problems. It is common, for instance, that the number of labeled samples is small but abundance of unlabeled samples is available or the number of labeled samples from different classes are extremely imbalanced. It is sometimes possible to use an already trained machine in a different domain. A machine could take advantage of external pieces of knowledge, in addition to labeled data to make more accurate predictions. In addition, not all real-world problems fit in a straightforward classification or regression problem, such as finding anomalies or outliers among data, especially streaming data. Besides aforementioned topics, this course will familiarize students with deep learning, reinforcement learning, multi-classifiers, genetic algorithms, and clustering textual documents. Among other evaluation criteria, this course entails a heavy experimental project. Offered by Info Sciences & Technology. May not be repeated for credit.