Machine Learning

Machine Learning

Machine learning – an area of artificial intelligence that focuses on using data to construct models to make predictions or decisions – is playing an important role in many areas of science. At Simula, our research focuses on the mathematical foundations of machine learning, the experimental study of machine learning algorithms, and the application of machine learning in real-life applications including sports, human health, and defense.  

"Machine learning, combined with the massive amounts of data and processing power that are available today, has enabled all sorts of applications, from fraud detection to virtual assistants to the translation of lost languages. There are many exciting advances ahead in areas such as precision medicine, autonomous vehicles and deep reasoning."

Sven-Arne Reinemo, Research Director

Departments

Research director

Focus areas

The amount of data produced in many domains has been increasing rapidly over recent years, but the tools and techniques to deal with the data are a work in progress. We develop machine learning methods that can be used in practical applications to process large amounts of data derived from multiple sources. Our focus areas are:

Developing techniques and mathematical foundations of machine learning.
A major challenge in dealing with large amounts of information is to combine (with data fusion techniques) and make sense of (with data mining techniques) different types of data from multiple sources, such as in biomedical research.

Dealing with such data is challenging not only because of the amount of data but also due to the varying characteristics of different data sources; some data (for example, an individual’s genetic makeup) is static while other variables might change over time, and some datasets might have unique components that are not shared by others. We develop multimodal deep learning techniques to handle this complexity. An important part of our work is performing calculations for, and optimisation of, tensor models, which can represent multisource heterogenous data.

Much of the data that we need to analyse is conveyed by signals that are monitored via sensor networks and another of our key focus areas is signal processing: investigating methods of extracting and processing information embedded in complex signals and images, such as video imagery.

Using machine learning in novel ways to solve real-world problems.
At its core, our work is about finding solutions for real-world scenarios. Our applied work includes:

  • training neural networks in object recognition to analyse images and video for use in domains including medicine and defence
  • developing data fusion and mining techniques to derive insights from data from multiple sources in the areas of sports and biomedical science
  • working with AI, computer vision and natural language processing to create realistic avatars to help in training social services professionals.

Spin-out companies

We are proud of the successful Oslo-based spinoff companies that have arisen from our work in machine learning at the Department of Holistic Systems (HOST).

Forzasys specialises in technology for sports analytics, namely: video streaming technology, distributed systems, networks and machine learning.

Augere creates technology based on video analytics and machine learning to analyse colonoscopy videos, helping doctors detect cancer.

Contract research opportunities in Machine Learning

We welcome opportunities to collaborate with government, industry and academic partners to develop machine-learning-based solutions to complex problems in areas such as health, sports and defence.

Selected projects

  • We are working with a consortia of European organisations and the European Defence Agency to develop an automatic threat detection and target recognition system, based on artificial intelligence and computer vision, that is faster than current manual processes and works in a range of environmental conditions. The system will be based on video-based analysis of human behaviour; sensor data fusion for defined combat distances based on 360° situational awareness; and real-time sensor information that is compared with historical data and integrated into the Command, Control, Communications, Computers and Intelligence (C4I) system.
  • We are collaborating on ILMA (Interview training of child-welfare and law-enforcement professionals interviewing maltreated children supported via artificial avatars), an interdisciplinary project funded by The Research Council of Norway and conducted with OsloMet Faculty of Social Sciences. In this project, we use AI, computer vision, and natural language processing to create realistic avatars to help train professionals in social services in the skills needed for interviewing children who have experienced abuse. The system uses data from past interviews with children to train avatars that can express emotion and respond spontaneously.
  • Improving sports performance with the Arctic University of Norway (UiT) and Forzasys AS at Tromsø’s Female Football Centre (FFC), which is funded by the Tromsø Research Foundation. We are developing non-invasive, privacy-preserving technology for sports science, using machine learning techniques that combine athlete-reported data with other information, including automated GPS and performance data. These tools enable us to continuously monitor athlete behaviour and can help predict performance and assist coaches in designing individualised programs.