I do research in machine learning and artificial intelligence, with frequent detours into the underlying fields of statistics, probability theory, and information theory. My Ph.D. supervisor is Csaba Szepesvári.
I belong to the Alberta Machine Intelligence Institute (AMII) and the Reinforcement Learning and Artificial Intelligence (RLAI) group.
See my list
of publications or my curriculum vitae for more information. I work on online learning and sequential
decision making, where an agent learns incrementally as it
continuously interacts with the world. In particular, I focus
on bandit problems and reinforcement
learning, both involving agents that learn to maximize a
reward which tells them how effective their actions are. I am currently working on reinforcement learning for complex
environments with many states. I am trying to understand
what makes such environments hard to plan in, and which
simplifying assumptions are needed to make them tractable. I
approach these questions through provably efficient
algorithms as well as theoretical bounds on
the hardness of planning in these environments. In Winter 2016, I was a principal instructor
for “Introduction
to Tangible Computing II” (CMPUT 275), an advanced
introductory course for computing science. In 2014 and 2015 I was
a teaching assistant for the same course. This course combines labs and lectures in a “studio” format with
three-hour sessions twice a week. Designed to Research Interests
Teaching
liberate you
from being a consumer of magic technology to a creator of it,
it includes object-oriented programming, the Python programming
language, and more complex algorithms and data structures such as
shortest paths in graphs; caching, memoization, and dynamic
programming; client-server style computing; recursion; and limited
distribution of computation tasks between the Arduino platform and
the traditional desktop in order to explore design tradeoffs.