On Learning Paradigms for the Travelling Salesman Problem

TL;DR

This paper investigates how different learning paradigms, specifically supervised and reinforcement learning, affect training neural networks to solve the Traveling Salesman Problem, highlighting RL's advantages in generalization and scale-invariance.

Contribution

It demonstrates that reinforcement learning outperforms supervised learning in training neural networks for the TSP, especially in generalizing to larger and variable graph sizes.

Findings

RL models generalize better to larger graphs

RL training does not require labeled data

RL produces scale-invariant solvers for new problems

Abstract

We explore the impact of learning paradigms on training deep neural networks for the Travelling Salesman Problem. We design controlled experiments to train supervised learning (SL) and reinforcement learning (RL) models on fixed graph sizes up to 100 nodes, and evaluate them on variable sized graphs up to 500 nodes. Beyond not needing labelled data, our results reveal favorable properties of RL over SL: RL training leads to better emergent generalization to variable graph sizes and is a key component for learning scale-invariant solvers for novel combinatorial problems.