ARC: Leveraging Compositional Representations for Cross-Problem Learning on VRPs
Han-Seul Jeong, Youngjoon Park, Hyungseok Song, Woohyung Lim
TL;DR
ARC introduces a compositional learning framework for cross-problem VRPs by disentangling attribute representations into intrinsic and contextual components. The intrinsic embedding is learned through analogical consistency to preserve invariant attribute semantics across problem variants, while the contextual embedding captures interactions conditioned on the intrinsic part. This results in strong generalization across in-distribution and unseen attribute combinations, plus effective few-shot adaptation and real-world applicability. The approach is validated through extensive experiments and analyses, establishing analogical embeddings as a potent tool for cross-problem learning in neural combinatorial optimization.
Abstract
Vehicle Routing Problems (VRPs) with diverse real-world attributes have driven recent interest in cross-problem learning approaches that efficiently generalize across problem variants. We propose ARC (Attribute Representation via Compositional Learning), a cross-problem learning framework that learns disentangled attribute representations by decomposing them into two complementary components: an Intrinsic Attribute Embedding (IAE) for invariant attribute semantics and a Contextual Interaction Embedding (CIE) for attribute-combination effects. This disentanglement is achieved by enforcing analogical consistency in the embedding space to ensure the semantic transformation of adding an attribute (e.g., a length constraint) remains invariant across different problem contexts. This enables our model to reuse invariant semantics across trained variants and construct representations for unseen combinations. ARC achieves state-of-the-art performance across in-distribution, zero-shot generalization, few-shot adaptation, and real-world benchmarks.