Currently I am interning at DeepMind with Vlad Mnih
My current research focus lie at the intersection between unsupervised representation learning, reinforcement learning, meta-learning and robotics. Specifically I am working on unsupervised representation learning via RL, and unsupervised RL.
In my past life I did atomic physics and quantum computing. You can see more of these work here.
Learning Plannable Representations with Causal InfoGANT Kurutach, A Tamar, G Yang, SJ Russell, P AbbeelWe propose an approach to learn plannable representations from high-dimensional sequential observations, where the generative process is induced by a transition in a low-dimensional planning model with additional noise