Machine Learning Research Group

We are a machine learning research group at Aalto University lead by Arno Solin. We are affiliated with the Department of Computer Science (School of Science at Aalto) and the Finnish Center for Artificial Intelligence (FCAI).

We focus on both theory and practice. Our goal is to develop new methods and algorithms that are able to adapt/learn from data but also obey structural or principal constraints. The research has a probabilistic (some say Bayesian) twist with a focus on uncertainty quantification and emphasis on scalable and even real-time inference. Even though we develop theory (methods and algorithms), we want to come up with things that actually work and can be shown to work even on limited hardware such as smart devices. This links the interest to information fusion where observations from different modalities (e.g., vision and movement) are combined.

Graduated Doctors

The following students have graduated from the group this far:

• Ari Heljakka (2020). "Deep Generative Neural Network Models for Capturing Complex Patterns in Visual Data". Supervisors: Juho Kannala and Arno Solin; Pre-examiners: Yingzhen Li (Microsoft Research Cambridge) and Dmitry Ulyanov (in3D); Opponent: Ole Winther (Technical University of Denmark).
• Santiago Cortés (2021). "Models and Methods for Inertial and Visual Odometry". Supervisors: Juho Kannala and Arno Solin; Pre-examiners: Yasutaka Furukawa (Simon Fraser University) and Isaac Skog (Linköping University); Opponent: Michael Blösch (Deep Mind).
• Joel Jaskari (2022). "Machine Learning for Healthcare". Supervisor: Arno Solin; Advisors: Simo Särkkä and Kimmo Kaski; Pre-examiners: Matthew Blaschko (KU Leuven) and Antti Airola (University of Turku); Opponent: Chris Holmes (University of Oxford).
• Yuxin Hou (2022). "Learning Latent Image Representations with Prior Knowledge". Supervisors: Juho Kannala and Arno Solin; Pre-examiners: Oisin Mac Aodha (University of Edinburgh) and Martin R. Oswald (University of Amsterdam); Opponent: Oisin Mac Aodha (University of Edinburgh).
• Lassi Meronen (2023). "Uncertainty Quantification in Deep Learning". Supervisor: Arno Solin; Pre-examiners: Carl Henrik Ek (University of Cambridge) and Vincent Fortuin (Helmhotz AI); Opponent: Henrik Ek (University of Cambridge). Received a best thesis award.

Funding and support

We acknowledge the computational resources provided by the Aalto Science-IT project and CSC, and support and help from NVIDIA AI Technology Center (NVAITC) Finland, Saab Finland, BusinessFinland, Helsinki Institute for Information Technology (HIIT), and the Finnish Center for Artificial Intelligence (FCAI).

We acknowledge funding from the Academy of Finland:

• Probabilistic principles for latent space exploration in deep learning
  (2021–2026, grant id 339730)
• Shallow models meet deep vision
  (2019–2023, grant id 324345)
• Sequential inference for real-time probabilistic modelling
  (2017–2020, grant id 308640)

How to join

The main route for hiring doctoral students is through the HICT call (twice yearly). Prospective doctoral students with strong background in stats/maths/ML can also contact Arno directly, but due to email load, I might not answer all emails.

Prospective post-doc candidates can contact Arno directly (with a CV, motivation letter, and research statement). Contracts are typically made for either one or two years. Positions are research-based and candidates are expected to have previous experience in publishing in relevant venues. We also participate in calls for post-docs in specific topic areas.

We hire summer interns and research assistants through the department call (local students) and the international Aalto ASci call. Arno also supervises M.Sc. theses at Aalto and occasionally also B.Sc. theses related to research topics in the group.