OATML

A paper by Yoav Gelberg, Tycho F.A. van der Ouderaa, Mark van der Wilk and Yarin Gal received the top award in the Proceedings track of the ICML Workshop on Geometry-grounded Representation Learning and Generative Modeling (GRaM). You can read the paper on variational inference failures under model symmetries here.

OATML group members Ilia Shumailov and Yarin Gal published a study in Nature which finds that indiscriminate use of model-generated content in training causes irreversible defects in the resulting models, in which tails of the original content distribution disappear. You can read the paper here.

OATML group members Sebastian Farquhar, Jannik Kossen and Yarin Gal, along with group alumni Lorenz Kuhn, published a study in Nature which introduces a breakthrough method for detecting hallucinations in LLMs using semantic entropy. You can read the paper here.

OATML student Gunshi Gupta was invited to present recent work along with collaborators from Georgia Tech and NYU (including group member Tim Rudner) at the Generative AI for Decision Making workshop (GenAI4DM) at ICLR 2024. The talk was on the paper ‘Pretrained Text-to-Image Diffusion Models Are Versatile Representation Learners for Control’. Read the paper here.

DPhil student Kelsey Doerksen will be a speaker at the Vice-Chancellor’s 2024 International Women’s Day Event, a discussion featuring panellists from across the University working in AI, who will share their views on making AI a force for gender equality and inclusion.

Together with Been Kim, Lisa Schut gave a talk at the Stanford HAI Fall Conference on New Horizons in Generative AI: Science, Creativity, and Society on ‘Leveraging AlphaZero to Improve our Understanding & Creativity in Chess’.

Yarin Gal has been appointed as the Director of Research at the new government Frontier AI Taskforce. As the Taskforce’s first progress report explains, Yarin’s appointment reflects his status as ‘a globally recognised leader in Machine Learning’.

DPhil student Jan Brauner has published an op-ed in TIME Magazine. He discusses the similarities between extinction risk from AI and present-day harms from AI.

OATML students Pascal Notin and Ruben Wietzman are co-organizing the Workshop on Computational Biology (WCB) at ICML 2023 jointly with collaborators at Harvard, Columbia, Cornell and others. OATML students Freddie Bickford Smith, Jan Brauner and Shreshth Malik are part of the program committee.

DPhil student Panagiotis Tigas and his collaborators from Frontiers Development Lab published their work on predicting geomagnetic pertubations using deep learning (DAGGER) in the Space Weather journal. The work has been featured in NASA’s main website

Senior Research Fellow Freddie Kalaitzis, in collaboration with a team from NASA Ames Research and the SETI institute, published a study in Nature Astronomy on a new deep-learning based framework to accelerate the search for life on Mars and other extra-terrestial environments. Freddie was the key AI contributor in this study. You can read the paper here.

OATML students Pascal Notin and Clare Lyle, along with OATML group leader Yarin Gal, are co-organizing the Machine Learning for Drug Discovery (MLDD) workshop at ICLR 2023 jointly with collaborators at GSK, Genentech, Harvard, MIT and others. OATML students Neil Band, Freddie Bickford Smith, Jan Brauner, Lars Holdijk, Andrew Jesson, Andreas Kirsch, Shreshth Malik, Lood van Niekirk and Ruben Wietzman are part of the program committee.

OATML graduate student Muhammed Razzak received a top reviewer award given to the top 8% of reviewers at NeurIPS 2022.

OATML PhD Student Kelsey Doerksen will join the Artificial Intelligence in Space: Benefits and Challenges panel at the Canadian Aeronautics and Space Institute ASTRO Conference.

OATML PhD Student Gunshi Gupta has been announced as one of the new Equality, Diversity and Inclusion fellows for the Mathematical, Physical & Life Sciences Division. Congratualtions to her, and all of the Fellows, who will help promote an inclusive and supportive culture for all.

OATML PhD Students Sören Mindermann, Muhammed Razzak, and Jan Brauner delivered a talk to Meta AI Research on their work on prioritised training.

OATML Senior Researcher Freddie Kalaitzis, along with Cristian Rossi and Maral Bayaraa, have published a white paper on the State of AI for Earth Observations: an intro to sensors, ML and applications. Read their work here.

Tim G. J. Rudner will join New York University as an Assistant Professor and Faculty Fellow at the Center for Data Science this fall.

OATML Senior Researcher Freddie Kalaitzis, in collaboration with WhyHow Ltd. and ESA Phi-lab, has released the WorldStrat dataset: the largest dataset of open high resolution satellite imagery with a CC-NC license, stratified by global land-uses, with an application to Multi-Frame Super-Resolution. Find the paper and code. Congratulations to the whole team involved!

OATML DPhil student Kelsey Doerksen will join the AI4SPACE: Perspective From the Next Generation panel at the International Astronautical Congress on September 19th, 2022, highlighting her research in Machine Learning for Space applications.

Prof. Gal, Jan Brauner, Sören Mindermann and Mrinank Sharma have won the MPLS Social Impact Award for their work on evaluating the effectiveness of Covid-19 internentions. Congratulations to all those awarded!

OATML DPhil student Pascal Notin will give a talk at the Center for Basic Machine Learning Research in Life Science (MLLS), ELLIS Unit in Copenhagen, on generative models for protein fitness and mutation effects prediction.

OATML PhD student Jan Brauner was a guest speaker at an event of the German Academic Scholarship Foundation (10th - 12th June 2022). He gave several talks and workshops on risks from advanced AI systems.

Freddie Kalaitzis will lead a €1M European Space Agency project to make Super Resolution technology for earth observations safer, more accountable, and open to everyone. Congratulations to the team!

OATML student Pascal Notin is co-organizing the 7th edition of the Workshop on Computational Biology (WCB) at ICML 2022 jointly with collaborators at Harvard, Columbia, Cornell and others. OATML students Neil Band, Freddie Bickford Smith, Jan Brauner, Andreas Kirsch and Lood van Niekirk are part of the PC.

OATML graduate students Lars Holdijk, Jannik Kossen, Clare Lyle, and Sören Mindermann are recognized as Highlighted Reviewers for their reviewing at ICLR 2022.

OATML DPhil student Panagiotis Tigkas will give an invited talk at Wellcome Centre of Human Neuroimaging, UCL on Bayesian Optimal Experimental Design for Causal Discovery.

OATML DPhil student Panagiotis Tigkas will give a talk at Machine Learning for Heliophysics conference on global geomagnetic perturbation forecasting using deep learning.

OATML graduate student Jannik Kossen receives a top reviewer award (top 10%) at AISTATS 2022.

OATML and GSK are co-organizing the GeneDisco Challenge at ICLR 2022. You can find out more about the challenge here, and read the GeneDisco paper here.

OATML DPhil student Kelsey Doerksen will give an invited talk as an expert speaker at Carleton University’s SOAR Student Leadership Conference.

OATML students Pascal Notin, Andrew Jesson and Clare Lyle, along with OATML group leader Professor Yarin Gal, are co-organizing the first Machine Learning for Drug Discovery (MLDD) workshop at ICLR 2022 jointly with collaborators at GSK, Harvard, MILA, MIT and others. OATML students Neil Band, Freddie Bickford Smith, Jan Brauner, Lars Holdijk, Andreas Kirsch, Jannik Kossen and Muhammed Razzak are part of the PC.

Professor Yarin Gal has been announced as one of five ‘Samsung AI Researcher of the Year’ award winners. The awards were launched last year to discover rising AI researchers globally.

The paper “Disease variant prediction with deep generative models of evolutionary data” was published in Nature. The work is a collaboration between OATML and the Marks lab at Harvard Medical School. It was led by Pascal Notin and Professor Yarin Gal from OATML together with Jonathan Frazer, Mafalda Dias, and Debbie Marks from the Marks lab. OATML DPhil student Aidan Gomez and Marks lab researchers Joseph Min and Kelly Brock are co-authors on the paper.

OATML graduate students Andreas Kirsch and Sebastian Farquhar received a reviewer award given to the top 8% of reviewers at NeurIPS 2021.

OATML DPhil student Kelsey Doerksen will be the manager of the 19th Space Generation Congress (SGC) hosted in Dubai, UAE in conjunction with the International Astronautical Congress.

Thirteen papers with OATML members accepted to NeurIPS 2021 main conference. More information in our blog post.

Jan Brauner will speak at the OECD Global Science Forum workshop on “Priority setting and coordination of research agendas: lessons learned from COVID 19”. The workshop will take place on October 4th and 5th, further information and a registration link can be found here.

Jan Brauner and Sören Mindermann have presented their work with Professor Yarin Gal and other collaborators on the effectiveness of mask-wearing at reducing COVID-19 transmission to the UK Cabinet Office and advised the office on mask-wearing policies.

Sebastian Farquhar has been elected a Junior Research Fellow in Computer Science at Christ Church, University of Oxford.

OATML graduate students Sebastian Farquhar and Jannik Kossen receive best reviewer awards (top 10%) at ICML 2021. Further, OATML graduate students Tim G. J. Rudner, Pascal Notin, Panagiotis Tigas, and Binxin Ru have served the conference as expert reviewers.

We’re co-organising the Bayesian Deep Learning Workshop at NeurIPS 2021 as well as two challenges: Approximate Inference in Bayesian Deep Learning and Shifts Challenge: Robustness and Uncertainty under Real-World Distributional Shift. This effort is led by Professor Yarin Gal, Neil Band, Sebastian Farquhar, and collaborators.

OATML graduate students Aidan Gomez, Jannik Kossen, and Neil Band will be presenting their recent paper Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning that introduces Non-Parametric Transformers at the Stanford Lecture Course ‘CS25: Transformers United’ on November 1, 2021. Professor Yarin Gal, Dr. Tom Rainforth, and OATML DPhil student Clare Lyle are co-authors on the paper.

The lecture is available online here.

OATML students Jannik Kossen and Neil Band will be presenting their recent paper Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning at Google Research on September 14, 2021. Professor Yarin Gal, Dr. Tom Rainforth, and OATML DPhil students Clare Lyle and Aidan Gomez are co-authors on the paper.

OATML MSc student Lood van Niekerk received a best poster award at the 2021 ICML Workshop on Computational Biology for his work on “Exploring the latent space of deep generative models: Applications to G-protein coupled receptors”. This is part of an ongoing collaboration between OATML and the Marks Lab. From OATML, Professor Yarin Gal and DPhil student Pascal Notin are co-authors on the paper.

OATML DPhil student Jannik Kossen gives invited talks at AI Campus Berlin on two recent papers: Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning and Active Testing: Sample-Efficient Model Evaluation. Recordings of are available upon request. Announcements are here and here. Professor Yarin Gal, Dr. Tom Rainforth, and OATML graduate students Sebastian Farquhar, Neil Band, Clare Lyle, and Aidan Gomez are co-authors on the papers.

OATML graduate student Lewis Smith and Professor Yarin Gal gave a talk on uncertainty in ML, discussing their collaboration with Adi Hanuka on uncertainty quantification for virtual diagnostics in the SLAC accelerator.

OATML research project has been covered by NASA, and featured on NASA’s home page. This research, published in Astronomy & Astrophysics, focuses on improving data for solar research. This was part of a collaboration between OATML and the Frontier Development Lab

Seven papers with OATML members accepted to ICML 2021, together with 14 workshop papers. More information in our blog post.

OATML students Jannik Kossen and Neil Band present their recent paper Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learning at Cohere on July 9, 2021. Professor Yarin Gal, Dr. Tom Rainforth, and OATML DPhil students Clare Lyle and Aidan Gomez are also co-authors on the paper.

OATML PhD Student Lewis Smith, along with Yarin Gal and Atılım Güneş Baydin, are part of the team invlved in developing an ML model to detect flood events, which has now been deployed in space. This is a project done in collaboration with the European Space Agency, Trillium Technologies and leaders in commercial AI, such as Google Cloud and Intel. You can read more about their work here

OATML graduate students Pascal Notin and Tim G. J. Rudner received best reviewer awards (top 5%) at UAI 2021.

Angelos Filos is one of the recipients of the 2021 J.P. Morgan PhD Fellowship. Congratulations to the winners!

Tim G. J. Rudner is one of 4 PhD students awarded the 2021 Qualcomm Innovation Fellowship Europe. Congratulations to the winners!

OATML graduate student Lewis Smith’s work together with Professor Yarin Gal, as part of FDL 2019, was published as a full length paper in Nature Scientific Reports. This work was a collaboration between the four members of his team, and academics from several universities.

OATML graduate student Pascal Notin will give an invited talk on Uncertainty in deep generative models with applications to genomics and drug design at the Cornell Machine Learning in Medicine seminar series. Professor Yarin Gal and collaborator José Miguel Hernández-Lobato are co-authors on the paper.

OATML graduate student Tim G. J. Rudner wrote a paper series on AI safety for non-experts (An Overview, Robustness, Interpretability).

OATML graduate student Pascal Notin will give an invited talk on “Large-scale clinical interpretation of genetic variants using evolutionary data and deep generative models” at the EMBL-EBI workshop on Machine Learning in Drug Discovery. The work is a collaboration between OATML (Pascal and Professor Yarin Gal) and the Marks Lab at Harvard.

OATML graduate student Seb Farquhar will speak on approximate Bayes in large neural networks drawing on Liberty or Depth (NeurIPS 2020) and Radial BNNs (AI Stats 2020) at the UCL Centre for Artificial Intelligence Seminar Series. Professor Yarin Gal, OATML graduate student Lewis Smith and collaborator Mike Osborne are co-authors on the papers.

Five papers with OATML members accepted to ICLR 2021:

• Invariant Representations for Reinforcement Learning without Reconstruction
• Improving Transformation Invariance in Contrastive Representation Learning
• On Statistical Bias In Active Learning: How and When to Fix It
• Improving VAEs’ Robustness to Adversarial Attack
• Capturing Label Characteristics in VAEs

OATML graduate student Tim G. J. Rudner will give an invited talk on Outcome-Driven Reinforcement Learning via Variational Inference at the MPI+UCLA Mathematical Machine Learning Seminar. Professor Yarin Gal and collaborator Sergey Levine are also co-authors on the paper.

OATML graduate student Tim G. J. Rudner will give an invited talk on “A Non-technical Guide to Modern Machine Learning” at the Georgetown University’s Center for Security & Emerging Technology.

OATML graduate student Tim G. J. Rudner will give an invited guest lecture on Bayesian Deep Learning at University College London.

The paper called “Inferring the effectiveness of government interventions against COVID-19” was published in Science today. The work is a collaboration with researchers from 9 universities, led by OATML graduate students Sören Mindermann and Jan Brauner, together with Mrinank Sharma from the Department of Statistics.

OATML hosted a joint workshop with Google Research on reliable machine learning with a series of talks and breakout sessions. The event was organized by OATML graduate student Tim G. J. Rudner and Mario Lučić at Google.

OATML Senior Research Fellow Freddie Kalaitzis will be giving a Spotlight Talk on Dynamic Hydrology Maps from Satellite-LiDAR Fusion at the AI for Earth Sciences workshop at NeurIPS 2020.

OATML graduate student Tim G. J. Rudner received a best reviewer award (top 10%) at NeurIPS 2020.

OATML graduate students Sören Mindermann and Jan Brauner, together with Mrinank Sharma from the Department of Statistics, were invited to give a talk about their work with Professor Yarin Gal on ‘inferring the effects of non-pharmaceutical interventions against COVID-19’, at the German Centre for Infection Research/University of Cologne.

OATML graduate student Tim G. J. Rudner will give an invited talk about his work with Professor Yarin Gal on Inter-domain Deep Gaussian Processes at the University of Cambridge’s ML@CS Seminar.

OATML DPhil student Angelos Filos is co-organising the NeurIPS 2020 Workshop on Talking to Strangers: Zero-Shot Emergent Communication.

OATML graduate student Lisa Schut presented alongside Rory McGrath, from Accenture Labs, on “Counterfactual Explanations: Making AI decision-making more useful and trustworthy.” The research presented was joint work with Oscar Key and Professor Yarin Gal, in collaboration with Accenture Labs.

OATML graduate student Tim G. J. Rudner has joined the OECD’s Working Groups on Trustworthy AI and AI Classification as an invited expert.

OATML graduate student Clare Lyle will be giving a talk about her work with Professor Yarin Gal on causal inference and generalization in deep reinforcement learning at the Simons Institute Workshop on Deep Reinforcement Learning on Thursday October 1.

Prof Yarin Gal will give a keynote talk at the Uncertainty for Safe Utilization of Machine Learning in Medical Imaging (UNSURE) workshop at the Medical Image Computing and Computer Assisted Intervention (MICCAI) conference. He will discuss work done at OATML on AI in medical imaging. The talk will cover work on topics ranging from data efficient AI to safe and interpretable AI.

Six papers with OATML members accepted to NeurIPS 2020:

• Liberty or Depth: Deep Bayesian Neural Nets Do Not Need Complex Weight Posterior Approximations
• Identifying Causal Effect Inference Failure with Uncertainty-Aware Models
• On the robustness of effectiveness estimation of nonpharmaceutical interventions against COVID-19 transmission
• Calibrating Deep Neural Networks using Focal Loss
• A Bayesian Perspective on Training Speed and Model Selection
• Neural Architecture Generator Optimization

Prof Yarin Gal’s MLSS seminars on Bayesian Deep Learning are available on YouTube (part 1 and part 2), with additional resources available here.

OATML graduate student Tim G. J. Rudner will give an invited talk on Inter-domain Deep Gaussian Processes at the UCL Centre for AI’s Statistical Machine Learning Seminar.

OATML graduate student Clare Lyle has been selected for the 2020 Open Philanthropy AI Fellowship. The fellowship will support her research for five years.

Five papers with OATML members accepted to ICML 2020:

• Invariant Causal Prediction for Block MDPs
• Uncertainty Estimation Using a Single Deep Deterministic Neural Network
• Inter-domain Deep Gaussian Processes
• Divide, Conquer, and Combine: a New Inference Strategy for Probabilistic Programs with Stochastic Support
• Can autonomous vehicles identify, recover from, and adapt to distribution shifts?

OATML graduate student Joost van Amersfoort was invited to talk about his work with Professor Yarin Gal on deterministic uncertainty quantification (DUQ) at Waymo (March 27th) and Rutherford Appleton Laboratory (June 18th).

OATML graduate students Jan Brauner and Sören Mindermann will give an invited talk at Africa CDC, Africa’s intercontinental public health agency, on June 12. They will present their work with Professor Yarin Gal on nonpharmacetical interventions against COVID-19 to the COVID-19 modelling group.

We’re organising the second “Quantification of Uncertainty in Segmentation” task as part of the Multimodal Brain Tumor Segmentation Challenge at MICCAI 2020, together with Angelos Filos, Raghav Mehta and Tal Arbel.

Our work with NASA and ESA over the past few years, together with Lewis Smith and Tim G. J. Rudner, is summarised in a recent article written by James Parr. Read more in Inspired: Bayesian deep learning for all humankind

Since 1999, MIT Technology Review has recognised young innovators and talented entrepreneurs from different countries who are developing new technologies to help solve the problems that affect our society. Yarin has been included in the ‘pioneers’ category. Read more: AI still makes mistakes, but his tools can alert us when a system is about to go wrong

We are glad to share 25 papers by OATML authors and collaborators presented at NeurIPS 2019 conference and workshops. Read more: blog post

Yarin Gal is one of five new Turing AI Fellows announced by The Alan Turing Institute. The Office for Artificial Intelligence, The Alan Turing Institute and UK Research and Innovation (UKRI) have worked together to successfully attract the Turing AI Fellows, some of the best research talent from around the world. Yarin will work on democratising safe and robust AI. Read more: Turing AI Fellows

Six group members honoured as top reviewers at NeurIPS 2019: 2 members among the top 400 highest scoring reviewers and awarded free registration (Tim G. J. Rudner and Sebastian Farquhar), and 4 among the top 50% reviewers (Zac Kenton, Andreas Kirsch, Angelos Filos and Joost van Amersfoort).

Five papers with OATML members accepted to NeurIPS 2019:

• BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning
• A Geometric Perspective on Optimal Representations for Reinforcement Learning
• VIREL: A Variational Inference Framework for Reinforcement Learning
• Variational Bayesian Optimal Experimental Design
• On the Benefits of Disentangled Representations

OATML DPhil student Angelos Filos is co-organising the NeurIPS 2019 Workshop on Emergent Communication: Towards Natural Language.

OATML graduate student Aidan Gomez will be presenting the Recurrent Neural Networks session with Kris Sankaran at Deep Learning Indaba on Tuesday 27 August. Schedule is available here and slides are available here.

We’re organising a “Quantification of Uncertainty in Segmentation” task as part of the Multimodal Brain Tumor Segmentation Challenge at MICCAI 2019, together with Angelos Filos, Raghav Mehta and Tal Arbel.

We’re organising the Fourth Bayesian Deep Learning Workshop at NeurIPS 2019. This effort is led by Professor Yarin Gal and collaborators.

OATML DPhil students Lewis Smith and Tim G. J. Rudner received ICML 2019 Outstanding Reviewer Awards (top 5% of reviewers).

Prof Yarin Gal will be speaking at the European Space Agency’s Phi-week, discussing work done at OATML on AI in space. The talk will cover work on topics ranging from exoplanet atmospheric retrieval to asteroid shape modelling from radar data.

OATML graduate student Angelos Filos will present his paper ‘Inverse Reinforcement Learning for Limit Order Book Dynamics’ at the the ICML 2019 Workshop on Applications and Infrastructure for Multi-Agent Learning, with the paper selected for an oral presentation.

Prof Yarin Gal will be speaking at the Royal Society’s Summer Science Exhibition on 6 July, discussing work on AI and art. The talk will cover techniques in machine learning that give us the chance to venture beyond the frame of some famous paintings.

OATML graduate student Aidan Gomez has been selected for the 2019 Open Philanthropy AI Fellowship. The fellowship will support his research for five years.

OATML DPhil student Aidan Gomez is co-organising the ICML 2019 Workshop on Invertible Neural Nets and Normalizing Flows together with Aaron Courville and Danilo Rezende.

Prof Yarin Gal will give a talk at the UN summit on “AI for Good” on the development of AI which we can trust: How can we develop trust-worthy machine learning tools to be deployed in the wild? He will further participate in a panel discussion on the topic IT’S A MATTER OF TRUST.

The Publications page has been updated with NeurIPS and UAI papers.

We’re organising the Third Bayesian Deep Learning Workshop at NeurIPS 2018. This effort is led by Professor Yarin Gal and collaborators.