ML4CO: Is GCNN All You Need? Graph Convolutional Neural Networks Produce Strong Baselines For Combinatorial Optimization Problems, If Tuned and Trained Properly, on Appropriate Data

Amin Banitalebi-Dehkordi,Yong Zhang

DOI: https://doi.org/10.48550/arXiv.2112.12251

2021-12-23

Abstract:The 2021 NeurIPS Machine Learning for Combinatorial Optimization (ML4CO) competition was designed with the goal of improving state-of-the-art combinatorial optimization solvers by replacing key heuristic components with machine learning models. The competition's main scientific question was the following: is machine learning a viable option for improving traditional combinatorial optimization solvers on specific problem distributions, when historical data is available? This was motivated by the fact that in many practical scenarios, the data changes only slightly between the repetitions of a combinatorial optimization problem, and this is an area where machine learning models are particularly powerful at. This paper summarizes the solution and lessons learned by the Huawei EI-OROAS team in the dual task of the competition. The submission of our team achieved the second place in the final ranking, with a very close distance to the first spot. In addition, our solution was ranked first consistently for several weekly leaderboard updates before the final evaluation. We provide insights gained from a large number of experiments, and argue that a simple Graph Convolutional Neural Network (GCNNs) can achieve state-of-the-art results if trained and tuned properly.

Machine Learning,Artificial Intelligence,Discrete Mathematics,Neural and Evolutionary Computing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is whether machine learning can be an effective option for improving traditional combinatorial optimization solvers under specific problem distributions, especially when historical data is available. Specifically, the paper explores whether, in the scenario of small data changes when repeatedly solving combinatorial optimization problems, machine - learning models can improve the performance of solvers by learning patterns in historical data. This was the core scientific problem in the 2021 NeurIPS Machine Learning for Combinatorial Optimization (ML4CO) competition, which aimed to improve the state - of - the - art combinatorial optimization solvers by replacing key heuristic components with machine - learning models. The Huawei EI - OROAS team achieved second place in the two - task of this competition and had been ranked first in multiple weekly leaderboard updates before the final evaluation. The team provided insights through a large number of experiments, demonstrating that simple Graph Convolutional Neural Networks (GCNNs) can achieve state - of - the - art results if properly trained and tuned on appropriate data.

ML4CO: Is GCNN All You Need? Graph Convolutional Neural Networks Produce Strong Baselines For Combinatorial Optimization Problems, If Tuned and Trained Properly, on Appropriate Data

The Machine Learning for Combinatorial Optimization Competition (ML4CO): Results and Insights.

Decision-focused Graph Neural Networks for Combinatorial Optimization

Learning to Solve Combinatorial Optimization under Positive Linear Constraints via Non-Autoregressive Neural Networks

Assessing and Enhancing Graph Neural Networks for Combinatorial Optimization: Novel Approaches and Application in Maximum Independent Set Problems

A Bi-Level Framework for Learning to Solve Combinatorial Optimization on Graphs

Towards a General Recipe for Combinatorial Optimization with Multi-Filter GNNs

Learning for Graph Matching and Related Combinatorial Optimization Problems

How Good is Neural Combinatorial Optimization? A Systematic Evaluation on the Traveling Salesman Problem

Graph Q-Learning for Combinatorial Optimization

Exact Combinatorial Optimization with Temporo-Attentional Graph Neural Networks

Towards General Algorithm Discovery for Combinatorial Optimization: Learning Symbolic Branching Policy from Bipartite Graph

Combinatorial optimization and reasoning with graph neural networks

Self-Improved Learning for Scalable Neural Combinatorial Optimization

Towards One-shot Neural Combinatorial Solvers: Theoretical and Empirical Notes on the Cardinality-Constrained Case

A Unified Pre-training and Adaptation Framework for Combinatorial Optimization on Graphs

Combinatorial Optimization with Automated Graph Neural Networks

Neural Combinatorial Optimization: a New Player in the Field

Scalable Primal Heuristics Using Graph Neural Networks for Combinatorial Optimization

Enhancing GNNs Performance on Combinatorial Optimization by Recurrent Feature Update