Northern lights over Finland. Photo by Maria Vojtovicova on Unsplash
The Alan Turing Institute (‘the Turing’) and The Finnish Centre for Artificial Intelligence (FCAI) signed a memorandum of understanding (MoU) in March 2019, formally establishing an ambitious partnership between the two organisations.
The FCAI – which conducts fundamental research on AI in cooperation with businesses and public sector organisations - is funded by the Academy of Finland. FCAI is Centre for Artificial Intelligence in Finland, initiated by Aalto University, University of Helsinki, and VTT Technical Research Centre of Finland. It aims to create "Real AI for Real People in the Real World"—a new type of AI, which can operate with humans in the complex world—and to renew Finnish industry with the latest AI.
The MoU has enabled both institutions to convene experts to work on shared research and translation projects. Although centred around the Turing’s data-centric engineering programme, a major research programme funded by the Lloyd's Register Foundation, interdisciplinary teams from domains including health have embarked on projects which support the Institute’s mission to use data and AI to pursue social good and address real-world issues through advancing world-class research. This has included the development of AI methods to improve the diagnosis of diabetic retinopathy - a project which is establishing one of the largest data collections of retinal images and optical coherence tomography (OCT) scans in the world. Diabetic retinopathy is a complication of diabetes caused by high blood sugar levels damaging the back of the eye (retina). It can cause blindness if left untreated. For more information: Uncertainty-Aware Deep Learning Methods for Robust Diabetic Retinopathy Classification.
There are a number of ongoing collaborations between FCAI and the Turing. Below, we present some highlights:
- Simo Sarkka (Aalto/FCAI), Toni Karvonen (Helsinki, previously the Turing), Chris Oates (Newcastle/Turing) and François-Xavier Briol (UCL/Turing) are developing novel probabilistic numerical methods. These methods use statistical tools to quantify the uncertainty underlying the numerical approximations underpinning machine learning algorithms. The work on this project has focused on methods for computing integrals through the Bayesian quadrature algorithm.
- Arto Klami (Helsinki/FCAI) Marcelo Hartmann (Helsinki/FCAI) and Mark Girolami (Cambridge/Turing) are developing sampling algorithms for Bayesian statistics by leveraging on differential geometry. Their goal is to provide algorithms that efficiently explore the true posterior distribution for complex models. See here for an example method.
- Sami Kaski (Aalto/FCAI, Manchester/Turing), Arto Klami (Helsinki/FCAI) and Theo Damoulas (Warwick/Turing) are developing the basis for human-AI collaboration in research, in the form of the Virtual Laboratories concept. Virtual Laboratories provide a new perspective on scientific knowledge generation and a means to incentivise new AI research into supporting research with tools that generalise across sciences. See here for an overview of Virtual Laboratories.
- François-Xavier Briol (UCL/Turing), Ayush Bharti (Aalto/FCAI) and Sami Kaski (Aalto/FCAI, Manchester/Turing) are developing novel methods for simulation-based inference. The aim of this project is to make these methods scalable to large engineering problems. The project focuses both on fundamental methodology, and applications in telecommunications engineering.
- Theo Damoulas (Warwick/Turing) and Arno Solin (Aalto/FCAI) are developing scalable approximations for Gaussian processes in the spatio-temporal setting.
- Kimmo Kaski and Theo Damoulas (Warwick/Turing) are developing robust Deep Learning methods to improve the diagnosis of diabetic retinopathy - a project which is establishing one of the largest data collections of retinal images and optical coherence tomography (OCT) scans in the world.
How to get involved
The Turing and FCAI run regular seminar series which anyone is free to attend. These should be helpful to identify some of the key research themes at each institute:
- Statistics in data-centric engineering seminar series at the Turing
- Machine learning coffee seminar at FCAI
Another way to start collaborations is to initiate your own projects, by for example organising a research visit. The following researchers are open to hosting researchers from FCAI or the Turing, as well as associated PhD students and postdoctoral researchers. Feel free to get in touch with them directly.
To apply for funding for such visits, you may want to apply for the FCAI visiting professor's programme. Funding for short-term visits (to and from FCAI) can be requested by FCAI professors using this form.
Related News and Events
- Turing-FCAI meetup
- Using machine learning to improve reliability of wireless communication system
- The Alan Turing Institute to work with The Finnish Centre for Artificial Intelligence on data science research
Dr Arto Klami, Helsinki University and Finnish Center for AI
Dr François-Xavier Briol, University College London and The Alan Turing Institute