A UK Spatial, Climate Health Probabilistic Programming Language Community

Project status

Ongoing

Introduction

For the UK to achieve global leadership in Probabilistic Programming Languages (PPL), it is necessary to cultivate a self-sustaining research community. Working collaboratively we will help realise this vision by supporting (1) conceptual advances in focal research areas of strategic importance to the UK (climate, health, and spatial modelling), (2) forecasting contests to build a community based around addressing policy needs, and (3) in-person and online training sessions. This work, with the full support of our three institutions, will leave a legacy of PPL leadership at the Turing beyond the lifespan of the call for submission that this project has been funded by.

Explaining the science

Our research is focused on the themes of space, climate and health and it will follow three objectives: 

1. Conceptual advances in climate changes, human health, and spatial modelling

In a world of epidemics accelerated by climate change and growing disease burden, there is an increasing demand for spatially fine-grained statistical models and estimates. While spatial modelling is a well-developed area within computational statistics, PPLs are not yet mature, robust, and scalable enough to meet the challenges of modern data analysis. Climate change models and the voluminous and ever-changing data they produce pose another huge computational burden on their own. Integrating them within PPLs is an important and open area of future research. We will address public policy and global health challenges and in order to do so, we need replicable data scientific workflows built on robust computational statistical methodologies, with PPLs playing a central role. PPLs are relatively unique in being able to naturally propagate uncertainties in data through into formally defined statistical probabilities of model correctness, which is a critical advantage when providing insights for policymakers. The UK, long a world leader in PPLs and small area statistics is uniquely positioned to catalyse a community around these challenges. 

2. Forecasting contests to support decision-makers

Communities build around common goals, and so we build a Probabilistic Programming Language (PPL) community around overcoming challenges identified at a Probabilistic Programming workshop held at the Turing in April 2023. We will organise annual workshops at the Turing to bring together researchers and data scientists at the intersections of PPL, spatial analysis, climate, and health. These workshops will identify policy-relevant questions that are challenging. For each research theme, we will organise an annual forecasting challenge. These forecasts will stimulate discussion and provide insights into policy-relevant questions.

3. Develop training resources to speed Probabilistic Programming Language (PPL) adoption

Our goal is to build PPL capacity, not just in terms of conceptual approaches and software advances, but also in terms of researchers’ capabilities. We will provide free, in-person ‘espresso courses’ at the Alan Turing Institute to engage developers, academics, and industry professionals with the theory and practicalities of working with.

Project aims

The aim of this project is to build a solid research community based around probabilistic programming language at the Alan Turing Institute. This will be achievable by making conceptual advances in climate changes, human health, spatial modelling, forecasting contests to support decision-makers and developing training resources to speed the Probabilistic Programming Language adoption.

Organisers

Researchers and collaborators