English

Stochastic Interpretation for the Arimoto Algorithm

Information Theory 2015-03-12 v2 math.IT

Abstract

The Arimoto algorithm computes the Gallager function maxQE0(ρ,Q)\max_Q {E}_{0}^{}(\rho,Q) for a given channel P(yx){P}_{}^{}(y \,|\, x) and parameter ρ\rho, by means of alternating maximization. Along the way, it generates a sequence of input distributions Q1(x){Q}_{1}^{}(x), Q2(x){Q}_{2}^{}(x), ... , that converges to the maximizing input Q(x){Q}_{}^{*}(x). We propose a stochastic interpretation for the Arimoto algorithm. We show that for a random (i.i.d.) codebook with a distribution Qk(x){Q}_{k}^{}(x), the next distribution Qk+1(x){Q}_{k+1}^{}(x) in the Arimoto algorithm is equal to the type (Q{Q}') of the feasible transmitted codeword that maximizes the conditional Gallager exponent (conditioned on a specific transmitted codeword type Q{Q}'). This interpretation is a first step toward finding a stochastic mechanism for on-line channel input adaptation.

Cite

@article{arxiv.1412.4510,
  title  = {Stochastic Interpretation for the Arimoto Algorithm},
  author = {Sergey Tridenski and Ram Zamir},
  journal= {arXiv preprint arXiv:1412.4510},
  year   = {2015}
}

Comments

5 pages, 1 figure, accepted for 2015 IEEE Information Theory Workshop, Jerusalem, Israel

R2 v1 2026-06-22T07:31:18.574Z