Research Interests

I am broadly interested in Learning and Adaptive systems which automatically acquire data, and continually improve their predictive and decision making capabilities. In this context, I am interested in the following questions:

Exploration vs Exploitation - spanning reinforcement learning, adaptive optimal control, and neuroscience. More generally, interactive learning settings with limited feedback interest me.
Deep Learning - Learning end-to-end systems that tightly couple perception and decision making (control) is facilitated by neural networks. I wish to study stochastic optimization and architectures for Deep RL, vision, and sequence prediction.
Common representation and multi-task learning - How can we effectively transfer knowledge across tasks and use multi-view (multi-modal) data. This might also be an effective way to perform regularization - since common representation is required to be informative for multiple tasks, it might not overfit for any task.

In addition to these questions, I am also keen to understand various matrix factorization problems with structural constraints (eg NMF, SPCA etc). This was my first encounter with ML, and I am interested in it's application to topic modeling, network reconstruction, matrix completion, and generalization to tensors. More generally, I am interested in any problem that has a mathematical (algorithmic) flavor and practical value.