Optimization and Convexity of log det(I+KX−1)
- 34 Downloads
This paper provides another proof for the convexity (strict convexity) of log det(I+KX-1) over the positive definite cone for any given positive semidefinite matrix K ⪰ 0 (positive definite matrix K ≻ 0) and the strict convexity of log det(K +X-1) over the positive definite cone for any given K ⪰ 0. Equivalent optimization representations with linear matrix inequalities (LMIs) for the functions log det(I+KX-1) and log det(K+X-1) are also presented. It was shown that these optimization representations with LMI constraints can be particularly useful for some related synthetic design problems. An iterative procedure based on the proposed LMI is presented to solve the minimax mutual information game with covariance and expected power constraints.
KeywordsCoding theory convex optimization log-det function matrix differential minimax mutual information game positive definite matrix
Unable to display preview. Download preview PDF.
- Y.-H. Kim and S.-J. Kim, “On the convexity of log det(I+ KX1),” arXiv preprint cs/0611043, 2006.Google Scholar
- N. J. Higham, Accuracy and Stability of Numerical Algorithms, Society for Industrial and Applied Mathematics, 2002.Google Scholar
- R. L. Dobrushin, “Optimum informtion transmission through a channel with unknown parameters,” Radio Eng. Electron., vol. 4, no. 12, pp. 1–8, 1959.Google Scholar
- T. M. Cover and J. A. Thomas, Elements of Information Theory, John Wiley & Sons, 2012.Google Scholar