Search papers, labs, and topics across Lattice.
This paper presents a self-supervised pretraining framework for graph combinatorial optimization, targeting routing problems like the Traveling Salesman Problem (TSP). By employing graph contrastive learning with geometric augmentations, such as rotations and axial reflections, the model learns invariant structural representations that enhance its performance. The hybrid pretraining approach yielded a significant 6.57% improvement in tour length for TSP1000, highlighting the importance of geometric pretraining as an inductive bias for scaling neural solvers to complex instances.
Geometric pretraining can lead to a 6.57% reduction in tour length for TSP1000, showcasing its potential to enhance neural solvers for complex routing problems.
This paper introduces a self-supervised pretraining framework for graph combinatorial optimization specifically designed to address the nature of routing problems like the Traveling Salesman Problem. By utilizing graph contrastive learning with geometric augmentations (specifically, rotations and axial reflections) the model is forced to learn invariant structural representations and global relative distance distributions. Results demonstrate that this pretraining strategy outperforms non-pretrained models across various problem scales. Notably, the hybrid strategy (combining rotation and reflection) achieved a 6.57% improvement in tour length for TSP1000, proving that geometric pretraining is an important inductive bias for effectively scaling neural solvers to high-dimensional instances.
