Advantage Actor Critic Benchmark

Proposed by Mnih et. al. in 2016 as A3C, A2C is the synchronous version of A3C (the extra "A" in A3C stands for "Asynchronous".

A2C is an extension of AC; the extra "A" stands for the Advantage function. We use GAE to compute for the loss; apart from that, the code is shared.