DeepACO: Neural-enhanced Ant Systems for Combinatorial Optimization
About
Ant Colony Optimization (ACO) is a meta-heuristic algorithm that has been successfully applied to various Combinatorial Optimization Problems (COPs). Traditionally, customizing ACO for a specific problem requires the expert design of knowledge-driven heuristics. In this paper, we propose DeepACO, a generic framework that leverages deep reinforcement learning to automate heuristic designs. DeepACO serves to strengthen the heuristic measures of existing ACO algorithms and dispense with laborious manual design in future ACO applications. As a neural-enhanced meta-heuristic, DeepACO consistently outperforms its ACO counterparts on eight COPs using a single neural architecture and a single set of hyperparameters. As a Neural Combinatorial Optimization method, DeepACO performs better than or on par with problem-specific methods on canonical routing problems. Our code is publicly available at https://github.com/henry-yeh/DeepACO.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Traveling Salesman Problem | TSPLib Large-scale instances | Objective Value7.81e+4 | 120 | |
| Traveling Salesman Problem | TSP-500 (test) | Gap1.84 | 85 | |
| Capacitated Vehicle Routing Problem | CVRP N=100 | Objective Value14.93 | 73 | |
| Traveling Salesman Problem | TSP50 | Optimality Gap0.00e+0 | 64 | |
| Traveling Salesman Problem | TSP-100 | -- | 56 | |
| Traveling Salesman Problem | TSP 1K (test) | Length23.85 | 45 | |
| Traveling Salesperson Problem | TSPLIB pr1002 | Optimality Gap20.96 | 36 | |
| Traveling Salesperson Problem | TSPLIB pr2392 | Optimality Gap (%)30.43 | 36 | |
| Traveling Salesman Problem | TSP N=20 | Optimality Gap2.23 | 33 | |
| Traveling Salesman Problem | TSP N=100 | Cost (%)0.00e+0 | 29 |