Achieved drastic improvement in sample complexity by using Kronecker Factored optimization in toy Reinforcement learning problems.

More details to be added soon...