Learning from A Single Graph is All You Need for Near-Shortest Path Routing
Primary Area: learning on graphs and other geometries & topologies
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: zero-shot learning, local search, reinforcement learning, all-pairs shortest path routing, network knowledge
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: We propose a simple algorithm that needs only a few data samples from a single graph for learning local routing policies that generalize across classes of geometric random graphs in Euclidean and hyperbolic metric spaces. We thus solve the all-pairs near-shortest path problem by training deep neural networks (DNNs) that let each graph node efficiently and scalably route (i.e., forward) packets by considering only the node’s state and the state of the neighboring nodes. Our algorithm design exploits network domain knowledge in the selection of input features and in the selection of a “seed graph” and its data samples. The leverage of domain knowledge provides theoretical assurance that the seed graph and node subsampling suffice for learning that is generalizable, scalable, and efficient. Remarkably, one of these DNNs we train —using distance as the only input feature— learns a policy that exactly matches the well-known Greedy Forwarding policy, which forwards packets to the neighbor with the shortest distance to the destination. We also learn a new policy, which we call Greedy Tensile routing —using both distance and stretch factor as the input features— that almost always outperforms greedy forwarding. We demonstrate the explainability and ultra-low latency runtime operation of Greedy Tensile routing by symbolically interpreting its DNN in terms as a low-complexity linear actions.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7891
Loading