News Archive

For the latest news, see the homepage.

[Mar 2025] We have just publicly released code and data for several of our recent works including implementations of the CABRA, GRLOS, and PRORL methods. Head to the Publications page for links to the relevant repositories.

[Feb 2025] Our paper A Cost-Aware Adaptive Bike Repositioning Agent Using Deep Reinforcement Learning has been published in IEEE Transactions on Intelligent Transportation Systems. We propose a reinforcement learning method for balancing resources in urban bike-sharing systems (BSS) that takes into account operational costs and is evaluated with real-world data from major cities. We demonstrate its advantages over existing techniques, including in a BSS with 1765 docking stations.

[Nov 2024] We presented our work Reinforcement Learning Discovers Efficient Decentralized Graph Path Search Strategies at the Third Learning on Graphs Conference (LoG 2024). Inspired by Milgram's small world experiment, we frame the problem of path search in graphs as a decentralized multi-agent decision-making process. We propose the GARDEN method and demonstrate its advantages over heuristic and learned baselines, including on real-world social media networks.

[Sep 2024] The paper Graph Reinforcement Learning for Combinatorial Optimization: A Survey and Unifying Perspective has been accepted for publication by TMLR. In this survey, we review works that have approached optimization problems over graphs with reinforcement learning techniques. Specifically, we focus on problems that currently do not have satisfactory exact or heuristic solutions, and for which RL can be advantageous as an approach for algorithm discovery.

[Aug 2024] I have joined the Goal-Oriented Autonomous Long-Lived Systems (GOALS) group at the Oxford Robotics Institute as a Postdoctoral Researcher. I am excited to continue my work on reinforcement learning and planning, dive into robotics applications, and develop graph learning techniques for autonomous systems.

[Aug 2024] Our work Trust-based Consensus in Multi-Agent Reinforcement Learning Systems was presented at the First Reinforcement Learning Conference (RLC). In this paper, we propose a trust mechanism for dealing with unreliable actors in decentralized multi-agent systems, and empirically show its effectiveness for solving a set of consensus environments. You can find a short video about the work here.

[Jul 2024] Our paper A Graph Reinforcement Learning framework for neural Adaptive Large Neighbourhood Search is now published in Computers & Operations Research. We propose a hybrid method that combines Graph RL and the ALNS metaheuristic, improving significantly on classic mechanisms as well as recently proposed RL techniques.

[May 2024] New pre-print: Large Language Models are Effective Priors for Causal Graph Discovery. This work studies the potential of leveraging LLMs as sources of prior information in the causal discovery process, highlighting their potential as well as shortcomings.

[Mar 2024] Our paper PRORL: Proactive Resource Orchestrator for Open RANs Using Deep Reinforcement Learning has been published in IEEE Transactions on Network and Service Management. We propose a reinforcement learning approach for dynamic allocation and orchestration of resources for the O-RAN infrastructure that underlies 5G communication technology.

[Nov 2023] New work: Tree Search in DAG Space with Model-based Reinforcement Learning for Causal Discovery. We address the problem of discovering causal graphs with a model-based reinforcement learning method, which is powered by an incremental algorithm for determining cycle-inducing edges, and is shown to compare favorably to model-free RL methods and greedy search.

[Apr 2023] Thrilled to have passed my PhD viva with no corrections for my dissertation "Learning to Optimise Networked Systems". I am grateful to my examiners Pietro Liò (University of Cambridge) and Simon Julier (UCL) for the stimulating conversation.

[Feb 2023] Our work Graph Reinforcement Learning for Operator Selection in the ALNS Metaheuristic was accepted for presentation at the International Conference in Optimization and Learning (OLA2023). We propose a hybrid method that combines RL and the ALNS metaheuristic, improving significantly on the operator selection mechanism in this classic method.

[Jan 2023] New work out in Proceedings of the Royal Society A: Planning spatial networks with Monte Carlo tree search. We propose a tree search framework for the construction of spatial networks. We improve in scalability over prior reinforcement learning methods, and perform case studies for improving the resilience and efficiency of Internet networks and metro systems.

[Jan 2023] Our paper RLQ: Workload Allocation With Reinforcement Learning in Distributed Queues has been published in IEEE Transactions on Parallel and Distributed Systems. We propose a scheduler for distributed task queues based on reinforcement learning. An implementation for the Celery framework in Python is available.

[Dec 2022] We presented our paper Dynamic Network Reconfiguration for Entropy Maximization using Deep Reinforcement Learning at the First Learning on Graphs Conference (LoG 2022). We propose a method for "scrambling" a network by increasing its entropy, and perform a case study for intrusion detection in cybersecurity.

[Sep 2022] Hot off the arXiv press: Graph Neural Modeling of Network Flows. In this work, we propose a graph neural network architecture for data-driven routing, that we evaluate on several ISP topologies. We also examine the relationships between graph structure and the difficulty of the routing task.

[Jul 2022] Excited to attend the Eastern European Machine Learning Summer School in Vilnius, Lithuania 🇱🇹.

[Oct 2021] Our paper Goal-directed graph construction using reinforcement learning has been published in Proceedings of the Royal Society A.

[Sep 2021] Our work Solving Graph-based Public Good Games with Tree Search and Imitation Learning was accepted for publication in NeurIPS 2021.