本文提出了一种结合强化学习（RL）和图嵌入的框架，用于自动学习解决图上NP-hard优化问题的贪心算法。通过不断构建解决方案并更新节点的图嵌入来训练评估函数Q，以最大化累积奖励。这种方法适用于各种图优化问题，与传统方法相比，它利用了问题的结构特征和分布。
Learning Combinatorial Optimization Algorithms over Graphs
Abstract
解决NP-hard问题通常需要大量的专业知识和反复试验, learn algorithm to automate this challenging, tedious process. learn heuristic algorithm利用结构特征解决问题. (same structure but differ in the data)的性质,
RL + graph embedding (incrementally constructs a solution)
how good the graph embedding network capturing the current state of the solution.决定了整个model的action .
该framework can be applied to a diverse range of optimization problems over graphs.
Introduction
传统解决np-hard graph optimization problem 又三种flavors:
- exact algorithms (enumeration or branch-and-bound) (缺点: difficult for large instance)
- polynomial-time approximation a
最低0.47元/天 解锁文章
2万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
