Mamba backbone powers new efficient neural combinatorial optimization framework

Researchers have developed ECO, an efficient framework for Neural Combinatorial Optimization that utilizes a Mamba backbone. This approach separates trajectory generation from gradient updates, employing a supervised warm-up phase followed by iterative Direct Preference Optimization on batched candidate sets. The framework incorporates a mixed Mamba encoder-decoder to manage memory growth and enhance hardware efficiency, alongside a local-search-guided bootstrapping strategy to stabilize training. ECO demonstrates superior performance, memory efficiency, and throughput on Traveling Salesperson Problem and Capacitated Vehicle Routing Problem benchmarks compared to existing neural baselines. AI

影响 Introduces a more memory-efficient and higher-throughput approach to neural combinatorial optimization, potentially impacting logistics and operations research.

排序理由 This is a research paper detailing a new framework and its performance on specific optimization problems.

在 arXiv cs.LG 阅读 →

• Direct Preference Optimization
• Mamba
• Zhenxing Xu
• Capacitated Vehicle Routing Problem

• 论文
• 基础设施

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →