Research map / Learning theory and algorithms / Combinatorial and search optimization
Neural Multi-Objective Combinatorial Optimization for Flexible Job Shop Scheduling Problems
Igor G. Smit, Yaoxin Wu, Pavel Troubil, Yingqian Zhang, Wim P.M. Nuijten
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add proposed reward formulation (bound-based) (ablation) | DCAN · FJSP 20x10 makespan-costs | hypervolume (greedy): 0.5182 -> 0.8083 | Improved |
| add lower-bound state features (ablation) | DCAN · FJSP 20x10 makespan-costs | hypervolume (greedy): 0.5068 -> 0.5182 | Improved |
| add lower-bound state features on top of reward (ablation) | DCAN · FJSP 20x10 tardiness-costs | hypervolume (greedy): 0.7948 -> 0.8112 | Improved |
| add proposed reward on top of state features (ablation) | DCAN · FJSP 20x10 tardiness-costs | hypervolume (greedy): 0.5734 -> 0.8112 | Improved |
| use conditional attention (DCAN) instead of weight-input WI-DAN | DCAN · FJSP 20x10 makespan-flowtime-costs | HV gap (greedy): 11.96% -> 7.12% | Improved |
| use DRL policy instead of MOEA/D | DCAN · FJSP 20x10 makespan-costs | HV gap: 42.93% -> 5.44% | Improved |
| use DCAN greedy instead of NSGA-II | DCAN · FJSP 20x10 makespan-flowtime-costs | HV gap: 42.01% -> 7.12% | Improved |
| use sampling (10 solutions per subproblem) instead of greedy | DCAN · FJSP 20x10 makespan-costs | HV gap: 5.44% -> 4.07% | Improved |
| increase number of subproblems from 10 to 105 (ablation) | DCAN greedy · FJSP 20x10 three-objective | HV gap: 12.41% -> 7.12% | Improved |
| increase number of subproblems from 105 to 1035 (ablation) | DCAN greedy · FJSP 20x10 three-objective | HV gap: 7.12% -> 4.55% | No clear effect |
| use DCAN on smallest instances instead of NSGA-II | DCAN · FJSP 10x5 makespan-flowtime-costs | HV gap: 10.02% -> 31.97% | Hurt |
| use multi-valued critic instead of single-valued critic | PPO critic · FJSP | performance | No clear effect |
Closest papers by meaning
- Neural Combinatorial Optimization for Stochastic Flexible Job Shop Scheduling Problems (AAAI 2025 · similarity 0.92)
- BOPO: Neural Combinatorial Optimization via Best-anchored and Objective-guided Preference Optimization (ICML 2025 · similarity 0.90)
- Learning-Guided Rolling Horizon Optimization for Long-Horizon Flexible Job-Shop Scheduling (ICLR 2025 · similarity 0.88)
- Learning Memory-Enhanced Improvement Heuristics for Flexible Job Shop Scheduling (NeurIPS 2025 · similarity 0.86)
- Improving Generalization of Neural Combinatorial Optimization for Vehicle Routing Problems via Test-Time Projection Learning (NeurIPS 2025 · similarity 0.86)
- CaDA: Cross-Problem Routing Solver with Constraint-Aware Dual-Attention (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
