Home / Research Interests / Publications / Prospectve Students
For Prospective Students
My group pursues research questions that demand rigor, hard work and cross-disciplinary exploration. Here are some of questions of interest.
- Large Scale Behvaior Adoption : How can we use social media to scale up cooperative behavior adoption to millions of people? This requires us to understand social networks as resource-constrained (people are busy, with limited time and money), with lossy communication (messages are missed, or read late) among members. Some interesting questions include: what is the cooperative capacity of such networks? what signaling schemes can be employed to achieve capacity? Can we nudge a network towards a behavior distribution? What can we learn from biological oscillators? Why isn't there a Nyqvist theorem for sampling behavior on graphs? Are there continuum alternatives to represent social interaction? I’m interested in theories, algorithms, applications and systems to support large-scale behavior adoption.
- Low-Power Algorithms: With the advent of mobile, untethered computing, saving power is important. I’m interested in exploring systematic approximation techniques to save power. Can we design algorithms that are correct 99% of the time, but save an order of magnitude of power?
- Can Friends help on a Date? : Today, Google and other search engines are excellent at locating the nearest Starbucks. They work with the implicit assumption that the queries are precise and the answers are unambiguous. They find, it hard, however, to answer questions like “best vegetarian restaurant for a date,” “should I buy the Nikon 80-200mm f2.8 lens?” because these questions require search engines to know a lot about us—our preferences and our interests. For many of these imprecise, high-level semantic queries, we rely on our friends, not Google. Unlike entering a factual query into Google, we rarely require these queries to be answered immediately. I’m interested in algorithms and systems that use our social networks to answer these questions, with some probabilistic guarantees on time and correctness.
- Machine Teaching: How do we design multimedia environments to optimally teach physical skills? This interest is exemplified by our work on stroke patient rehabilitation. What strategy should we adopt to minimize the time taken to teach a particular stroke patient, a specific physical skill (e.g. to tie a shoe-lace), while ensuring high quality of movement? How do we efficiently explore the space of different learning strategies for a skill? I’m interested in algorithms that adapt these immersive environments to teach skills.
I’m interested in hearing from self-motivated students who are intellectually curious, unafraid to fail and who are keen to apply their knowledge to address real-world problems. If you are interested in these questions and you plan to get in touch with me, be sure to first read some of the papers published by my group.