Scholarlink

In this paper, we examine possibilities for global optimization on the GPU using interval-based libraries. We compare a purely brute-force based approach, a monotonicity test based approach and a Sivia based approach on the GPU with each other and with popular CPU tools in C-XSC and Octave. For that, we employ a problem in the context of optical multiple-input multiple-output (MIMO) systems. We demonstrate that, although the relatively simple minimization problem we consider is difficult to solve by means of general global optimization, GPU-based methods are able to provide a solution, which is even verified if GPU implementations of the underlying libraries are verified.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GPU-Based Interval Optimization in the Context of Optical MIMO Systems

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ekaterina Auer,
  • Andreas Ahrens,
  • Lorenz Gillner

摘要

In this paper, we examine possibilities for global optimization on the GPU using interval-based libraries. We compare a purely brute-force based approach, a monotonicity test based approach and a Sivia based approach on the GPU with each other and with popular CPU tools in C-XSC and Octave. For that, we employ a problem in the context of optical multiple-input multiple-output (MIMO) systems. We demonstrate that, although the relatively simple minimization problem we consider is difficult to solve by means of general global optimization, GPU-based methods are able to provide a solution, which is even verified if GPU implementations of the underlying libraries are verified.