作者: Karthick N. Parashar , Olivier Sentieys , Daniel Menard
关键词:
摘要: Trading off accuracy to the system costs is popularly addressed as word-length optimization (WLO) problem. Owing its NP-hard nature, this problem solved using combinatorial heuristics. In paper, a novel approach taken by relaxing integer constraints on variables and obtain an alternate noise-budgeting This uses quantization noise power introduced into due fixed-point word-lengths instead of actual valued word-lengths. The proved be convex in rounding mode case can therefore analytical solvers. An algorithm with linear time complexity provided order realize from budgets obtained solving