Marcus Liwicki is a Chair Professor and Head of Subject for Machine Learning at Luleå University of Technology. Professor Liwicki joined WASP in 2022 when his research group became one of two WASP Affiliated Groups of Excellence at Luleå University of Technology.
What is your position/role in WASP?
Head of the Machine Learning group at Luleå University of Technology (affiliated group of excellence).
Why did you choose to join WASP?
Great networking opportunity, joint research, and collaboration projects.
What are the benefits you see in WASP?
Good opportunity to work on basic research projects towards the aims of WASP. Interesting presentations during the conferences and study trips for our PhD students.
Briefly describe your research topic.
- Machine Learning for the welfare of the society
- Machine Learning methods for few data
- Deep representation learning
- Applied AI in general
For years we studied “questioning” the standard engineering approaches applied in deep learning. Back in 2015, we found that one can easily fool CNN with just fixing one pixel; later we showed that auto-encoders are not always useful for pre-training (better would be PCA or LDA), and recently we have shown ways to make contrastive learning more powerful (on data different than natural images and also for implementations on limited hardware). See for example:
- Prakash Chandra Chhipa, Richa Upadhyay, Gustav Grund Pihlgren, Rajkumar Saini, Seiichi Uchida, and Marcus Liwicki. Magnification Prior: A Self-Supervised Method for Learning Representations on Breast Cancer Histopathological Images. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023, pp. 2717-2727.
- Chhipa, P.C. et al. (2023). Depth Contrast: Self-supervised Pretraining on 3DPM Images for Mining Material Classification. In: Karlinsky, L., Michaeli, T., Nishino, K. (eds) Computer Vision – ECCV 2022 Workshops. ECCV 2022. Lecture Notes in Computer Science, vol 13807. Springer, Cham.
In what way can your research be of importance to our society in the future?
- Sustainable machine learning (energy efficient and available for everyone)
- Equal benefits from machine learning for everyone (especially for people with handicap)
- Secure machine learning