Welcome and Introduction (Carsten Welsch / University of Liverpool, Julie Sheldon / LJMU, Adam Ruby / KPMG Liverpool)

• Julie Sheldon (LJMU)
• Carsten Welsch (University of Liverpool)
• Adam Ruby (KPMG Liverpool)

AgenticAI for Data Analysis (Boris Bolliet / Cambridge University)

• Boris Bolliet (Cambridge University)

Multi-agent systems consisting of Large Language Model (LLM) and Retrieval Augmented Generation (RAG) powered assistants are game-changing for data analysis tasks. Compared to a standard script that would run a pipeline from A to Z, we can interact dynamically with intelligent agents to ask for modifications, or to request more information on the analysis at hand while the pipeline is being developed and executed. This way, large portions of data analysis pipelines become automated in a fully controlled manner. We will show examples where we set-up widely used codes in CMB analyses such as CAMB and CLASS to perform calculations and execute them within minutes while the same would take hours for a human. This illustrates first small steps towards fully AI assisted cosmological data analysis.

(Machine) Learning to create artwork and quantum fields (Pavel Buividovich / University of Liverpool)

• Pavel Buividovich (University of Liverpool)

Machine Learning algorithms for image generation are often considered as competitors for human artistic activities. We discuss a somewhat unexpected aspect of this competition, demonstrating how GenAI can help to detect human-made artwork forgeries made in the style of famous painters. We then discuss a less visual, but technically more advanced application of GenAI to the generation of "snapshots" (configurations) of quantum fields. In contrast to image generation, where success is measured by human perception, this application of GenAI imposes much stricter constraints on statistical properties of the output data which motivate deeper mathematical insights in GenAI models. We review some of the state-of-the-art GenAI algorithms suitable for generation of both artistic images and quantum fields.

Self supervised Learning in Astrophysics (Anna Scaife / University of Manchester) I will review why self-supervised learning is so important for astronomy, what the current most popular approaches are for self-supervised learning and how these are being applied to astronomical data. This lecture will recap some of the general principles of deep-learning, but forward reading is advised for those who are less familiar with the topic.

Open source science and making an impact (James Nightingale / Newcastle University)

• James Nightingale (Newcastle University)

Open source science applies the principles of open source software development to scientific research, emphasizing transparency, collaboration, and accessibility to make scientific knowledge and data freely available. I will argue that adopting these practices to the highest standard is crucial for the advancement of science. Through case studies from the “reproducibility crisis”, I will highlight the potentially devastating consequences of not practicing open science. I will then demonstrate how projects like SciPy, NumPy, and Pandas have transformed the research landscape. Open source science requires significant time, effort, and energy, and may not always be rewarded by the current scientific funding landscape. Nevertheless, I will contend that it ultimately enhances your research output, productivity, and is a compelling means to build collaboration with commercial industry partners.

Bioinformatics and the Curse of Dimensionality (Euan McDonnell / University of Liverpool)

• Euan McDonnell (University of Liverpool)

Graph Neural Nets, Application of AI to Sports (Zhe Wang / DeepMind)

• Zhe Wang (DeepMind)

Identifying key patterns of tactics implemented by rival teams, and developing effective responses, lies at the heart of modern football. However, doing so algorithmically remains an open research challenge. To address this unmet need, we propose TacticAI, an AI football tactics assistant developed and evaluated in close collaboration with domain experts from Liverpool FC. We focus on analysing corner kicks, as they offer coaches the most direct opportunities for interventions and improvements. TacticAI incorporates both a predictive and a generative component, allowing the coaches to effectively sample and explore alternative player setups for each corner kick routine and to select those with the highest predicted likelihood of success. We validate TacticAI on a number of relevant benchmark tasks: predicting receivers and shot attempts and recommending player position adjustments. The utility of TacticAI is validated by a qualitative study conducted with football domain experts at Liverpool FC. We show that TacticAI’s model suggestions are not only indistinguishable from real tactics, but also favoured over existing tactics 90% of the time, and that TacticAI offers an effective corner kick retrieval system. TacticAI achieves these results despite the limited availability of gold-standard data, achieving data efficiency through geometric deep learning.

Prize ceremony and wrap up