Research

The MUPI group carries research on statistical machine learning and artificial intelligence.

Our main goals are to

  1. Develop computational techniques for learning under uncertainty, based on approximate Bayesian inference and probabilistic programming
  2. Solve interesting data-driven applications, focusing in problems short on data that hence need advanced modeling techniques based on data integration, semi-supervised learning etc.

Most of our current main activities fall under the first three topics (Probabilistic inference, AI for ultrasonics, and Virtual Laboratories) with active ongoing projects, but we still work also on the other topics listed here.

Probabilistic inference

Statistical machine learning provides tools for understanding complex data collections, using Bayesian inference to cope with uncertainty on model parameters originating from learning from finite data. We develop computationally efficient and maximally automatic approximate algorithms for Bayesian inference, in context of probabilistic programming and machine learning. Our goal is to allow the user to focus on model specification, not needing to worry about the specifics of inference.

Highlights:

  1. Efficient Riemannian Monte Carlo methods based on a metric with computationally efficient inverse (Hartmann et al. Lagrangian manifold Monte Carlo on Monge patches, AISTATS 2022; Yu et al. Scalable Stochastic Gradient Riemannian Langevin Dynamics in Non-Diagonal Metrics, TMLR 2023).
  2. Riemannian Laplace approximation (Yu et al. Riemannian Laplace approximation with the Fisher metric, AISTATS 2024).
  3. Integration of approximate posterior inference and decision-making, by calibrating variational approximations to bettere account for the eventual decision task (Kuśmierczyk et al. Variational Bayesian Decision-making for Continuous Utilities, NeurIPS'19) and by post-hoc calibtration with flexible decision-making modules (Kuśmierczyk et al. Correcting Predictions for Approximate Bayesian Inference, AAAI 2020).
  4. Review on current state of prior elicitation with recommendations for future directions (Mikkola et al. 
    Prior Knowledge Elicitation: The Past, Present, and Future, Bayesian Analysis 2024).
  5. Using prior predictive distribution for expert knowledge elicitation (Hartmann et al. Tackling uncertainty in prior predictive elicitation, UAI 2020) and for fast hyperparameter optimization of hierarchical Bayesian models (da Silva et al. Prior specification for Bayesian matrix factorization via prior predictive matching, JMLR 2023).
  6. Bayesian inference under combinatorial constraints. (Klami and Jitta. Probabilistic size-constrained microclustering, UAI 2016)

Current projects:

  1. Efficient Riemannian Inference (ERI, 2022-2024). Development of efficient geometric MCMC algorithms for Bayesian inference. Funded by Research Council of Finland ICT-2023 program.
  2. Computationally efficient inference on Riemann embedding manifolds (CORE, 2022-2025). Research Council of Finland postdoctoral researcher project of Marcelo Hartmann.
  3. Finnish Center for Artificial Intelligence (FCAI), Agile probabilistic AI research theme.

Past projects:

  1. Scalable probabilistic analytics (SPA, 2016-2018). Development of computationally efficient variational inference algorihms for probabilistic programs. Funded by Tekes, Reaktor, Ekahau and M-Brain. Collaboration with Petri Myllymäki and Teemu Roos.
  2. Reliable Automatic Bayesian Machine Learning (RAB-ML, 2018-2019). Development of reliable and efficient solutions for Bayesian machine learning and probabilistic programming. Funded by Research Council of Finland ICT-2023 program. Consortium with Aki Vehtari and Antti Honkela.
AI for ultrasonics

We build machine learning and artificial intelligence tools for modeling ultrasound propagation in complex environments. We develop methods e.g. for inverse problems (detecting fouling or deformations), focusing ultrasound for cleaning, and acoustic levitation. The work is done is done together with the group of Ari Salmi and Edward Haeggström working on ultrasound physics, and we also collaborate with Altum Technologies that provides practical ultrasonic cleaning solutions.

The main activities are currently aiming for sustainable and safe cleaning of industrial production equipment, for reducing the environmental and economical harm of fouling that accumulates over time. We develop AI-enhanced sensing technologies for detecting and quantifying the fouling and for controlling the cleaning process so that the risk of damage is minimized.  

Highlights:

  1. First example of AI model for detection and monitoring of fouling during ultrasonic sensing, based on monotonicity. (Rajani et al. Detecting industrial fouling by monotonicity during ultrasonic cleaning, MLSP'18).
  2. Detection and fouling quantification with Gaussian processes using line integral observations. (Sillanpää et al. Ultrasonic Fouling Detector Powered by Machine Learning, IEEE International Ultrasonics Symposium (IUS), 2019; Longi et al. Sensor Placement for Spatial Gaussian Processes with Integral Observations, Uncertainty in Artificial Intelligence (UAI) 2020).
  3. Neural networks for localization of internal structure with chaotic cavity. (Sillanpää et al. Localizing a target inside an enclosed cylinder with a single chaotic cavity transducer augmented with supervised machine learning, AIP Advances, 2021).

Current projects:

  1. Sustainable industrial ultrasonic cleaning (SIUC, 2023-2025). Ensuring the sensing and cleaning technologies can be used in sustainable and safe manner. The project has received funding from the European Union (NextGenerationEU instrument) and is funded by the Research Council of Finland. Consortium with Ari Salmi.

Past projects:

  1. Machine learning for ultrasonic cleaning (ML-UC, 2019-2023). Development of machine learning and artificial intelligence solutions for modeling ultrasound propagation in fouled structures. Funded by Research Council of Finland.
  2. Ultrasonic AI-powered Industrial Sensing Platform (UA-ISP, 2021-2023). Development of AI-driven distributed ultrasonic sensing platform for low-cost holistic monitoring of complex environments. Funded by  Research Council of Finland ICT 2023 program. Consortium with Ari Salmi.
Virtual Laboratories

Virtual Laboratories are a new perspective to scientific knowledge generation. Many of the elements in scientific discovery are general across scientific domains, and by isolating them from domain-specific elements (models, simulations, theories) we can develop AI techniques for assisting scientific discovery as well as industrial R&D more efficiently. Any research environment, for instance a natural science laboratory, can leverage on these techniques by framing their operations as a virtual laboratory.

Highlights:

  1. Klami et al. Virtual Laboratories: Transforming research with AI provides a high-level vision of the concept.
  2. VAI-LAB, a proof-of-concept open source tool for setting up virtual laboratories

Current projects:

  1. Virtual Laboratories for pharmaceutical R&D (Business Finland Co-Research project, 2023-2025). Research towards establishing Virtual Laboratories in pharmaceutical research, in collaboration with research partners Samuel Kaski (Aalto), Markus Heinonen (Aalto) and Luigi Acerbi (Univ. of Helsinki) and industrial partners Orion, Silo AI, Pharmatest Services, Admescope and CSC.
  2. Finnish Center for Artificial Intelligence FCAI
Human behavior

Today it is easy to collect information about individuals by monitoring their activities, either based on explicit sensors or by log data collected by computers they interact with. We develop models required for inferring interesting and useful information based on such data, to describe, understand and enchance our daily life.

Highlights:

  1. Modelling risk behavior of individuals from observed data in conctext of computer games. (Tanskanen et al. Modeling Risky Choices in Unknown Environments, ACML'21)
  2. Models for learning personalised effect of interactions (personalised treatment effect of uplift) from highly unbalanced data collections. (Nyberg et al. Uplift Modeling with High Class Imbalance, ACML'21)
  3. Modeling keyboard usage in programming education and other educational contexts. (Leinonen et al. Automatic inference of programming performance and experience from typing patterns, ACM TSCSE'16)
  4. Human computer interaction in touch-based information retrieval. (Andolina et al. Intentstreams: smart parallel search streams for branching exploratory search, IUI'15)
  5. Understanding intentions based on brain signal analysis, in particular MEG recordings during natural tasks. (Kauppi et al. Towards brain-activity-controlled information retrieval: Decoding image relevance from MEG signals, NeuroImage 2015)

Past projects:

  1. Traces of Information: Intelligence from Fragmented Sources (ToI, 2013-2019). Academy Research Fellow project of Arto Klami.
  2. Machine Insight for Behavioral Analytics (MINERAL, 2019-2022). Business Finland Co-Innovation project. Consortium with Antti Oulasvirta (Aalto University) and companies.
Data integration

Machine learning research is often carried out in the context of elegant but simplified setups: It is assumed that all relevant data is provided in form a simple matrix or tensor. In most practical applications this is not the case, but instead we need to combine information scattered in multiple data sources of heterogeneous nature. We provide fundamental modeling solutions for combining such data sources.

Highlights:

  1. Bayesian canonical correlation analysis and inter-battery factor analysis (IBFA) and their extension Group factor analysis (GFA) for discovering relationships between more than two parallel data sources. (Klami, Virtanen and Kaski. Bayesian canonical correlation analysis, JMLR, 2013; Klami et al. Group factor analysis, IEEE Transactions on Neural Networks and Learning Systems, 2015).
  2. Matrix factorization tools for discovering relationships between more complex setups and heterogeneous data. (Klami, Bouchard and Tripathi. Group-sparse embeddings in collective matrix factorization, ICLR 2014; Klami. Polya-gamma augmentations for factor models, ACML 2015)
  3. Cross-domain object matching for discovering relationships between data sources with no known pairing between objects. (Jitta and Klami. Few-to-few cross-domain object matching, Advanced methodologies for Bayesian networks, 2017; Klami. Bayesian object matching, Machine Learning, 2013).

Past projects:

  1. Traces of Information: Intelligence from Fragmented Sources (ToI, 2013-2019). Academy Research Fellow project of Arto Klami.
  2. Improved Learning by Combinin Information Sources (ILCIS, 2013-2015). Funded by Xerox Research Foundation, collaboration with Abhishek Tripathi (Xerox Research Center India) and Petri Myllymäki.
Hyperspectral imaging

Hyperspectral cameras capture the full spectrum of light, instead of jus the three channels of red, green and blue that mimic the limited vision of humans. Having access to this richer information makes most computer vision problems easier. The existing HS cameras are, however, expensive and large. We develop a low-cost alternative that uses AI to process images captured with a passive add-on device that can be attached to any camera, brining HS imaging to smartphones and DSLRs. We also work on hyperspectral image analysis.

Highlights:

  1. Deep learning algorithms for hyperspectral image acquisition (Toivonen et al. Snapshot hyperspectral imaging using wide dilation networks, Machine Vision and Applications, 2020).
  2. Methods and applications for hyperspectral image interpretation: Luotamo et al. Multi-scale Cloud Detection in Remote Sensing Images using a Dual Convolutional Neural Network, IEEE Transactions in Geoscience and Remote Sensing, 2020 and Toivonen et al. Epiphyte colonisation of fog nets in montane forests of the Taita Hills, Kenya, Annales Botanici Fennici, 2020.

Past projects:

  1. Mobile hyperspectral imaging and computer vision platform (2019-2021). Development of methods and algorithms for acquisition of hyperspectral images with mobile devices, their use for computer vision, and preparation for commercialization of the result. Funded by Business Finland under the New Business from Research Ideas (TUTLI) instrument.
Data-efficient modeling

Many modern machine learning models are complex and require large training data sets, which makes learning difficult in applications where labeling examples is costly or difficult. We study data-efficient techniques for learning complex models from limited supervision, by utilizing related learning tasks (multi-task learning, transfer learning) and unlabeled observations or external constraints (semi-supervised learning). We also develop solutions for changing enrivonments based on domain adaptation.

Highlights:

  1. Structured pseudo-labels for semi-supervised learning based on output-space constraints or smoothness assumptions. (Longi, Pulkkinen and Klami. Semi-supervised convolutional neural networks for identifying Wi-Fi interference sources, ACML'17).
  2. Transfer learning in computer science education. (Lagus et al. Transfer learning methods in programming course outcome prediction. ACM Transactions on Computer Education, 2018:18(4)).
  3. Segmentation of large multispectral satellite images based on coarse annotations in limited-memory computational architecture. (Luotamo et al. Multi-scale Cloud Detection in Remote Sensing Images using a Dual Convolutional Neural Network, IEEE Transactions in Geoscience and Remote Sensing, 2020).

Past projects:

  1. Scalable probabilistic analytics (SPA, 2016-2018). Development of computationally efficient variational inference algorithms for probabilistic programs. Funded by Tekes, Reaktor, Ekahau and M-Brain. Collaboration with Petri Myllymäki and Teemu Roos.