notesum.ai

Published at November 29

RL-MILP Solver: A Reinforcement Learning Approach for Solving Mixed-Integer Linear Programs with Graph Neural Networks

cs.LG

cs.AI

Released Date: November 29, 2024

Authors: Tae-Hoon Lee¹, Min-Soo Kim¹

Aff.: ¹Korea Advanced Institute of Science and Technology

Arxiv: http://arxiv.org/pdf/2411.19517v1

Refer to caption

Dataset	Method	Metric
Dataset	Method	FR $\uparrow$	FFST $\downarrow$	RP_Gap $\downarrow$
$D_{9*18}$	Gurobi	100/100	$<1$	0
	M_MLP	100/100	0.05	4.45
	M_CNN	100/100	0.05	1.90
	Ours	100/100	0.07	0
$D_{100*50}$	Gurobi	100/100	$<1$	0.01
	M_MLP	100/100	0.06	1.60
	M_CNN	100/100	0.07	1.96
	Ours	100/100	0.06	1.33