Traditionally machine learning problems are catgorized as either supervised or unsupervised. In an unsupervised
problem like clustering, the goal is to look for patterns from raw data without any human input. In supervised learning
the goal is to build accurate predictors using limited human input in the form of training data. However, technologies such as crowdsouricng
platforms, wearable gadgets, and social media are making it much easier for humans to provide constant feedback and interact with learning
algorithms in more sophisticated ways. This motivates the study of new models and algorithms for interactive learning. In this talk
I will describe recent progress and challenges in designing theoretical models and algorithms for interactive learning in the
context of clustering and classification problems.