Talks and events

Whilst Bayesian deep learning (BDL) promises to address reliability and uncertainty in AI by combining Bayesian methods with deep learning, fundamental challenges and misconceptions have limited its practical impact, calling for a critical reassessment of BDL’s role and potential compared to non-Bayesian alternatives in real-world applications.

I’ll be attending and giving a talk at the Adaptive Experimentation Workshop organised by Meta.

Step-DAD: Semi-Amortized Policy-Based Bayesian Experimental Design was accepted at the Data-centric Machine Learning Research (DMLR) ICLR Workshop.

Invited to give a talk at the Challenges in Inversion and Model Calibration: From UQ to OED minisymposium.

Guest speaker for Statistical Machine Learning course at UCL

Review of the paper LLaMA: Open and Efficient Foundation Language Models by Touvron et al. (2023)

Efficient Real-world Testing of Causal Decision Making via Bayesian Experimental Design for Contextual Optimisation (with J. Jenning, C. Zhang and A. Foster), poster presenation at ICML 2022 Workshop on Adaptive Experimental Design and Active Learning in the Real World.

Women in Machine Learning (WiML) un-workshop poster session, where I presented some of my work on Bayesian optimal experimental design in real-time applications.

I have been selected as a finalist in the Mathematics category and will be presenting some of my research in the Houses of Commons.

This house believes that AI is better, stronger, faster at picking stocks for investment. The motion will be proposed by Ed Fishwick, Managing Director, Global Co-Head of Risk & Quantitative Analysis at BlackRock, and opposed by me.

Presented our Deep Adaptive Design (DAD) method at the Un-Conference track of the OxML summer school.

Review of the paper “Nonmyopic Multifidelity Active Search” by Nguyen et al. (2021), which introduces the MF-ENS algorithm.

Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design (with A. Foster, I. Malik and T. Rainforth), long presentation at ICML 2021, Bayesian Learning 1 session.

Discussion of our recent paper Deep Adaptive Design (with A. Foster, I. Malik and T. Rainforth).

Review of the paper “Playing Atari with Deep Reinforcement Learning” by Mnih et al (2013) which introduced deep Q-networks (DQN).