Research Program 2 (R2)
Simulator-based inference
The goal of FCAI’s Research Program Simulator-based inference is to develop methodology for the new AI having efficient, interpretable reasoning capability, by cross-breeding modern machine learning and simulator-based inference. Simulator-based inference contributes to mainly FCAI research objectives Data efficiency (objective 1) and Understandability (objective 3).
Current research in Simulator-based inference includes Engine for Likelihood-free Inference (ELFI) software, which builds a community-driven ecosystem of simulator models and inference algorithms. The new method has accelerated inference by several orders of magnitude. The other main branch of this Research Program includes groundbreaking work on simulator-based deep learning and generative adversarial networks (GANs).
The advances in high-dimensional models and causal AI-based reasoning will make inference in human-AI interaction possible using cognitive models, in areas such as personalized medicine (e.g., cancer therapy assistance to clinicians), widely deployed R&D tools for chemists, engineers, and physicists (e.g., for materials design), and modeling for multiple other applications by enabling significantly more complex models based even on limited data.
Examples of publications:
Jukka Corander, William P Hanage, Johan Pensar. 2022. Causal discovery for the microbiome. Lancet Microbe.
Joakim Löfgren, Dmitry Tarasov, Taru Koitto, Patrick Rinke, Mikhail Balakshin, Milica Todorović. 2022. Machine Learning Optimization of Lignin Properties in Green Biorefineries. ACS Sustainable Chem. Eng. 10, 9469.
Tero Karras, Miika Aittala, Samuli Laine, Erik Härkönen, Janne Hellsten, Jaakko Lehtinen, Timo Aila. 2021. Alias-Free Generative Adversarial Networks. Advances in Neural Information Processing Systems (Proc. NeurIPS).
Caroline Colijn, Jukka Corander, Nicholas J. Croucher. 2020. Designing ecologically optimized pneumococcal vaccines using population genomics. Nature Microbiology, 5.
Tero Karras, Miika Aittala, Janne Hellsten, Samuli Laine, Jaakko Lehtinen, Timo Aila. 2020. Training Generative Adversarial Networks with Limited Data. Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS), Advances in Neural Information Processing Systems, 33.
Coordinating professor: Jukka Corander – jukka.corander at helsinki.fi
People
The groups of following professors take part in the Research Program Simulator-based inference. If you would like to join this program, please contact the coordinating professor.
Luigi Acerbi, University of Helsinki
Jukka Corander, University of Helsinki, University of Oslo and Sanger Institute – coordinating professor
Antti Honkela, University of Helsinki
Perttu Hämäläinen, Aalto University
Samuel Kaski, Aalto University
Arto Klami, University of Helsinki
Ville Kyrki, Aalto University
Jaakko Lehtinen, Aalto University, NVIDIA
Antti Oulasvirta, Aalto University
Teemu Roos, University of Helsinki
Kai Puolamäki, University of Helsinki
Simo Särkkä, Aalto University
Aki Vehtari, Aalto University
Fundamental AI Research
Joint methodological goal
AI-assisted decision-making, design and modeling →
Research Programs
Probabilistic AI →
Simulators →
Deep learning →
Privacy and security →
Interactive AI →
Autonomous AI →
AI in society →
Highlight Programs
Modeling tools →
Health →
Service assistant →
Atmospheric →
Materials →
Sustainability →