FSC with Feedback and Causal State at the Encoder
In this talk we review a general finite-state channel (FSC) with the state causally available to the encoder and with output feedback. The FSC family is rich and includes, for instance, a channel with Markovian state, where the state has memory and is input-independent, and also cases where the state is input-dependent such as the energy harvesting model. We derive a new achievable rate expression for this general setting and show that it can be formulated as an infinite-horizon average-reward Markov decision process (MDP). The general expression turns out to achieve the capacity of two known unifilar FSC channels with feedback, namely the trapdoor channel and the input-constrained binary erasure channel. For the noiseless Binary Energy Harvesting Channel (BEHC), it provides rates comparable to existing achievable rates. The expression is also computable for the noisy BEHC with feedback, and to our best knowledge this is the first achievable rate given for such a setting.
The talk is based on joint work with Oron Sabag and Prof. Haim Permuter.
Eli just finished his Electrical and Computer Engineering B.Sc degree with honor (with emphasis on telecommunications and signal processing) from Ben-Gurion University, Israel. He is currently starting his Ph.D. studies in the direct track program, under the supervision of Prof. Haim Permuter. He does research in information theory that involves machine learning techniques such as reinforcement learning.