News
[October, 2024] Delighted to announce that four papers led by my PhD students, spanning topics from LLM alignment to foundational models, have been accepted to NeurIPS 2024.
[June, 2024] Talk at SecondMind
[November, 2023] Talks at ICARL (Imperial College London), INSAIT and Google DeepMind (London)
[September 7, 2023] New paper “Distributionally Robust Model-based Reinforcement Learning with Large State Spaces” is now available on arXiv.
[September 5, 2023] Excited to announce that I have been honored with the EPSRC New Investigator Award to continue our work on robust decision making and reinforcement learning.
[July 6, 2023] Our NeurIPS 2023 workshop proposal submission, Adaptive Experimental Design and Active Learning in the Real World, has been accepted to the conference [NeurIPS 2023 Workshops].
[July 6, 2023] Our NeurIPS 2023 workshop proposal submission, New Frontiers of AI for Drug Discovery and Development, has been accepted to the conference [NeurIPS 2023 Workshops].
[June 12, 2023] Delighted to announce that I have been honored with the Google Research Scholar Program award for Machine Learning & Data Mining. News item: ICCS Researcher receives Google Research Scholar Program Award
[March 14, 2023] Talk at Huawei London.
[January 21, 2023] Our paper “Near-optimal Policy Identification in Active Reinforcement Learning” got accepted to ICLR 2023 as notable-top-5% (oral)!
[January 21, 2023] Our paper “Efficient Planning in Combinatorial Action Spaces with Applications to Cooperative Multi-Agent Reinforcement Learning” got accepted to AISTATS 2023.
[January 12, 2023] We started an online reading group on modern adaptive experimental design and active learning in the real world.
[November 11, 2022] Talk at UCL (DeepMind/ELLIS CSML Seminar).
[October 28, 2022] Talk at Oxford Uni. (AIMS seminar).
[October 21, 2022] Talk at Imperial College London.
[October 4, 2022] Talk at Google Brain Seminar.
[September 16, 2022] Three papers got accepted to NeurIPS 2022:
- Graph Neural Network Bandits
- A Robust Phased Elimination Algorithm for Corruption-Tolerant Gaussian Process Bandits
- Movement Penalized Bayesian Optimization with Application to Wind Energy Systems
Publications
Right Now, Wrong Then: Non-Stationary Direct Preference Optimization under Preference Drift
Adversarial Robust Decision Transformer: Enhancing Robustness of RvS via Minimax Returns-to-go
Group Robust Preference Optimization in Reward-free RLHF
REDUCR: Robust Data Downsampling Using Class Priority Reweighting
Distributionally Robust Model-based Reinforcement Learning with Large State Spaces
Safe Model-Based Multi-Agent Mean-Field Reinforcement Learning
Efficient Planning in Combinatorial Action Spaces with Applications to Cooperative Multi-Agent Reinforcement Learning
Near-optimal Policy Identification in Active Reinforcement Learning
Movement Penalized Bayesian Optimization with Application to Wind Energy Systems
Graph Neural Network Bandits
A Robust Phased Elimination Algorithm for Corruption-Tolerant Gaussian Process Bandits
Misspecified Gaussian Process Bandit Optimization
Risk-averse Heteroscedastic Bayesian Optimization
Efficient Model-Based Multi-Agent Mean-Field Reinforcement Learning
Online Submodular Resource Allocation with Applications to Rebalancing Shared Mobility Systems
Combining Pessimism with Optimism for Robust and Efficient Model-Based Deep Reinforcement Learning
DP-Sniper: Black-Box Discovery of Differential Privacy Violations using Classifiers
Stochastic Linear Bandits Robust to Adversarial Attacks
Contextual Games: Multi-Agent Learning with Side Information
Learning to Play Sequential Games versus Unknown Opponents
Corruption-Tolerant Gaussian Process Bandit Optimization
Mixed Strategies for Robust Optimization of Unknown Objectives
Distributionally Robust Bayesian Optimization
No-Regret Learning in Unknown Games with Correlated Payoffs
Overlapping Multi-Bandit Best Arm Identification
Robust Adaptive Decision Making: Bayesian Optimization and Beyond
Adversarially Robust Optimization with Gaussian Processes
High Dimensional Bayesian Optimization via Additive Models with Overlapping Groups
Robust Maximization of Non-Submodular Objectives
Streaming Robust Submodular Maximization: A Partitioned Thresholding Approach
A Distributed Algorithm for Partitioned Robust Submodular Maximization
Robust Submodular Maximization: A Non-Uniform Partitioning Approach
Lower Bounds on Regret for Noisy Gaussian Process Bandit Optimization
An Efficient Streaming Algorithm for the Submodular Cover Problem
Truncated Variance Reduction: A Unified Approach to Bayesian Optimization and Level-Set Estimation
Time-Varying Gaussian Process Bandit Optimization
Learning-Based Compressive Subsampling
Active Learning of Self-concordant like Multi-index Functions
Near-Optimally Teaching the Crowd to Classify
