University of California, Santa Cruz
This paper studies a random coding ensemble in which each codeword is constrained to satisfy multiple cost constraints. Using this ensemble, an achievable second-order coding rate is presented for the mismatched single-user channel. Achievable error exponents are given for the mismatched single-user channel and the matched multiple-access channel. The method of random coding is a ubiquitous tool in information theory for proving the existence of codes with a vanishing error probability. In particular, the i.i.d. random-coding ensemble introduced by Shannon has found use in a vast range of settings. An alternative ensemble is the constant-composition ensemble in which each codeword has the same empirical distribution. In many settings, this ensemble yields performance gains over the i.i.d. ensemble.