On the discrete superposition model of partially cognitive interference channels

We find an achievable rate region for the partially cognitive interference channel in the weak interference regime, where the cognitive transmitter learns the primary message over a noiseless link with finite capacity. The rate region is established in the discrete superposition model, which has the property of approximating the AWGN model within a constant gap. In the high SNR regime our region is close to known outer bounds on the AWGN capacity region, yet its formulation is sufficiently simple to give valuable insights. Interestingly, the cognitive transmitter does not need to learn the complete primary message in order to achieve the rate region of the fully cognitive interference channel. The capacity of the noiseless link between the two transmitters is sufficiently large if it equals the single-user capacity of the cross-link from primary transmitter to secondary receiver.