Date/heure
Date(s) - 30 janvier 2015

Catégories Pas de Catégories

$Summary\; :\; A\; new\; splitting\; algorithm\; is\; proposed\; to\; minimize\; the\; sum\; of\backslash nconvex\; functions,\; potentially\; nonsmooth\; and\; composed\; with\; linear\backslash noperators.\; This\; generic\; formulation\; encompasses\; numerous\; regularized\backslash ninverse\; problems\; in\; image\; processing.\; The\; algorithm,\; whose\; weak\backslash nconvergence\; is\; proved,\; calls\; the\; individual\; gradient\; or\; proximity\backslash noperators\; of\; the\; functions,\; without\; any\; inner\; loop\; or\; linear\; system\; to\backslash nsolve.\; The\; classical\; Douglas-Rachford,\; forward-backward\; and\backslash nChambolle-Pock\; algorithms\; are\; recovered\; as\; particular\; cases.\; In\; the\backslash nsecond\; part\; of\; the\; talk\; (joint\; work\; with\; A.\; Hirabayashi,\; Kyoto,\; Japan),\backslash nI\; address\; the\; problem\; of\; recovering\; a\; spike\; train\; from\; noisy\; lowpass\backslash nmeasurements,\; through\; a\; reformulation\; as\; a\; structured\; low\; rank\; matrix\backslash napproximation\; problem.\; Used\; as\; a\; heuristic\; for\; this\; nonconvex\; problem,\backslash nthe\; proposed\; algorithm\; yields\; state-of-the-art\; results.[$