The final report of BigInsight

Here is the final report of BigInsight (2015—2024), which was one of the third generation Norwegian Centres for Research-based Innovation. It was funded by the Research Council of Norway and by fifteen partners.

Final report

summary

Objectives and vision: When we started in 2015, BigInsight was the first AI centre in Norway. We did not call it AI at the time, but machine learning: “BigInsight’s objective is to develop methods, algorithms and computational tools in model based statistics and machine learning to solve innovation challenges at our partners and in science and society.” More specifically, we identified two broad, yet precise challenges: personalised solutions and prediction of transient phenomena. These two areas have developed in these nine years to become main pillars of contemporary machine learning.

“We identified two broad, yet precise challenges: personalised solutions and prediction of transient phenomena.”

Consortium and categories of partners: The consortium consists of one research institute (NR, host), two universities (UiO and UiB), three public research organisations (OUS, NIPH, Cancer Registry of Norway), three public service partners (NAV, Skatteetaten, Statistics Norway), and six private companies (DNB, Gjensidige, ABB, DNV, Telenor, Norsk Hydro). Of these, Statistics Norway joined BigInsight in 2018, while all the others were part of the centre from day one. All the partners are large organisations and key players in the Norwegian economy and welfare state.

“We quickly adapted to the rapidly changing research front, where Explainable AI and applying deep learning for new domains are two examples which were important both in terms of innovation and research.”

Scientific results: The scientific footprint of BigInsight is substantial. We have published more than 200 scientific papers and contributed to important new statistical and machine learning methods for example for anomaly detection and monitoring of vessels, personalised computer simulations of breast cancer treatment for tailored treatments, machine learning for prediction of mortgage defaults, Explainable AI taking feature dependence into account, and modelling of COVID-19 in Norway. We have become internationally renowned in Explainable AI – a field that did not exist when BigInsight started in 2015. The foundation for the COVID-19 modelling was laid in BigInsight before the pandemic, and during the pandemic we conducted real time research, where the innovation took place immediately, while the research was published later. BigInsight also produced a considerable number of open software packages. 

“When the COVID-19 pandemic hit us, BigInsight was ready and within a few weeks dedicated lots of its resources to this monumental challenge for Norway.”

Results and impact for industry, public sector and society at large: All partners have had a keen interest in the development of new or improved processes or services, building up a strengthened knowledge base, improved access to competent personnel and research institutions, and recruitment of qualified personnel. Some partners that also make use of NR’s contracted research have pointed out that strengthening NR’s competence and foundation is in their interest as well as in NR’s interest. Possibly somewhat surprising, many of the partners really appreciated the informal exchange of knowledge and ideas between private and public actors, and the improved network it provided. From the annual BigInsight day, hosted by the user partners, to the weekly BigInsight lunches as well as other informal or organised meetings across the centre set the stage for this. In the report, we present some of our success stories, which include “Benefits for the maritime industry”, “First in Norway to request authorization from the Norwegian Data Protection Authority for Machine Learning research on personal data“, and “The COVID-19 story”. BigInsight’s greatest achievement and societal impact was the modelling of COVID-19 during the pandemic in Norway. 

Researcher training and master’s level education: BigInsight has significantly impacted research and education at the UiO, fostering a thriving data science community. This impact is evident in the establishment of the annual Data Science Day, which attracts 500 attendees, and a new master’s program in data science, a joint venture between the Departments of Mathematics and Informatics. The centre also influenced the mathematics curriculum, introducing courses focused on big data and machine learning, reflecting the growing importance of these fields. Over 45 master's students completed theses with BigInsight, benefiting from joint supervision by university staff, NR researchers, and partner representatives, strengthening connections between these groups. BigInsight supported numerous PhD students, with 26 successfully defending their theses by February 2025, demonstrating its commitment to developing future researchers. The centre also hosted over 30 postdoctoral researchers, contributing to both academia, with some securing permanent positions, and industry, as others transitioned to business roles. Finally, BigInsight's activities contributed to the establishment of dScience, a centre supporting computational and data science research at UiO.

International cooperation: BigInsight's international collaborations have led to several EU funded projects: RESCUER and BD4QoL, both of which utilize data-driven methods to improve breast cancer and head and neck cancer treatments respectively, and ENFIELD, European AI Lighthouse focused on promoting adaptive, green, and trustworthy AI. BigInsight has also played a significant role in the capacity development of higher education in Ethiopia. Besides, BigInsight has had partnerships with STOR-i, University of Lancaster, The Medical Research Council Biostatistics Unit, University of Cambridge, and the Department of Mathematics, University of Minneapolis. Further, BigInsight was a founding partner of a Nordforsk project aimed at creating a joint Nordic long-term academic collaboration on pandemic preparedness using advanced mathematical modelling and systematically collected health data.

Added value of organising the activities as a centre: BigInsight’s results would have been much less pronounced if it had not been organised as a joint centre for research and user partners. The size made it possible to take more risk, divert resources to new ventures, such as Explainable AI, set the stage for formal and informal cooperation across the centre, and, maybe most importantly, be an inspiring place to work.

Legacy of the centre: The centre lives partly on through Integreat, a Centre of Excellence funded by the Research Council of Norway from 2023 with UiO, UiT (The Arctic University of Norway) and NR as partners. A large consortium around BigInsight has applied for a Research Centre for Artificial Intelligence, funded by The Research Council of Norway. This centre – called TRUST – will be approximately twice as big as BigInsight. The funding will be assigned in June 2025.

Lars Henry Berge Olsen from BigInsight successfully defended his thesis

BigInsight congratulates Lars Henry Berge Olsen from BigInsight and the Department of Mathematics, who successfully defended his thesis "What’s in the Black Box? Improving the Shapley Value Explanation Methodology" for the degree of Philosophiae Doctor on February 7, 2025.

Main research findings:

Artificial intelligence and machine learning models are increasingly used in predictive tasks across various sectors, e.g., finance, healthcare, and retail, due to their high-performance accuracy. However, their predictions are often impossible to understand due to their opaque decision-making processes. To address this issue and comply with regulations like the EU AI Act, explainable AI (XAI) has emerged to explain how and why the model reached specific predictions. A prominent XAI framework is Shapley values, which originated in cooperative game theory in the 1950s but is today extensively used as a tool that can explain predictions made by any model.

This thesis improves the Shapley value explanation methodology in four key areas: developing methods to incorporate feature dependencies into the explanations, making them more precise and trustworthy; categorizing both existing and novel methods into distinct classes for thorough comparison and evaluation, with accompanying user recommendations; introducing stabilization and correction strategies to improve the accuracy of Shapley value approximations consistently; and creating the open-source shapr package for R and Python to ensure that these advancements are accessible to researchers and practitioners.

Adjudication committee:

  • Professor Marvin Wright, University of Bremen 

  • Associate professor Inga Strümke, Norwegian University of Science and Technology 

  • Associate professor Johan Pensar, University of Oslo 

Supervisors:

  • Professor Ingrid Kristine Glad, University of Oslo

  • Dr. Martin Jullum, Norwegian Computing Center

  • Dr. Philos Kjersti Aas, Norwegian Computing Center

 

Lars Henry Berge Olsen

Ingrid Dæhlen from BigInsight successfully defended her thesis

BigInsight congratulates Ingrid Dæhlen from BigInsight and the Department of Mathematics, who successfully defended her thesis "The best of the bad: Precise and robust estimation and model selection" for the degree of Philosophiae Doctor on January 31, 2025.

Main research findings:

A model is a mathematical description of some phenomenon. Models can help us predict, understand and shape the world around us. That being said, the world is complex, and the language of mathematics is rigid. This gives models a major flaw. They are almost always wrong. But models work well in practice and are crucial for humans to make sense of the world. This thesis attempts to guide scientists in the search for good models, by deriving ways of evaluating, fitting and ranking candidates without trusting them blindly. This is achieved by formulating precise approximations to the "error" of models, and by studying methods for fitting wrong models in the best possible way.

Adjudication committee:

  • Professor Ingrid van Keilegom, University of Leuven

  • Associate Professor Kristoffer Herland Hellton, OsloMet

  • Associate Professor David Ruiz Banos, University of Oslo

Supervisors:

  • Associate Professor Ingrid Hobæk Haff, Universitetet i Oslo

  • Professor Nils Lid Hjort, Universitetet i Oslo

Ingrid Dæhlen

SIMON BOGE BRANT FROM BIGINSIGHT SUCCESSFULLY DEFENDED HIS THESIS

BigInsight congratulates Simon Boge Brant and the department of Mathematics, who successfully defended his thesis “Selected topics in regression with a binary outcome: Fraud detection, and applications of copulas to logistic regression” for the degree of Philosophiae Doctor on November 8th, 2024.

Main research findings:

The thesis concerns different topics that all are related to logistic regression - models for the log-odds where inference is based on a (Bernoulli) likelihood for the conditional distribution of the outcome given covariates. Three papers make up the thesis. The first of these discusses a fraud detection problem, which we summarise in terms of a loss function, and discuss strategies for finding suitable models that approximately minimise this function. In both the second and third papers we present new models, and algorithms for fitting these. In the first of these we construct a boosting algorithm where the base learners are copula-based regression models (Noh et al., 2013; Chang and Joe, 2019), and in the last paper we present an extension of the linear logistic regression model, which is constructed through an R-vine based extension of a discriminant analysis analogue of the linear logistic regression model.

Adjudication committee:

  • Professor Thomas Nagler, Ludwig Maximilian University of Munich 

  • Professor Bart Baesens, KU Leuven 

  • Professor Geir Storvik, University of Oslo

 Supervisors:

  • Associate Professor Ingrid Hobæk Haff, University of Oslo

  • Associate Professor Riccardo De Bin, University of Oslo

Simon Boge Brant

Nye videoer ute om XAI-verktøyet eXplego

Har du laget en maskinlæringsmodell og sliter med å skjønne hva modellen egentlig legger vekt på? eXplego er et enkelt verktøy som hjelper deg som data scientist eller KI-utvikler med å velge en forklaringsmetode som faktisk gir deg svar på det du vil forklare. Verktøyet består av et interaktivt trediagram, der du blir guidet frem til en forklaringsmetode ved å svare på spørsmål om ditt forklaringsbehov.

 Det finnes en versjon på norsk og en litt mer teknisk versjon på engelsk.

Videoene er av: Jacob Sjødin (NAV), Martin Jullum (NR), Robindra Prabhu (NAV), Wivi Eilertsen (NAV) og Tore Græsdal (NAV). Produsent: Arild Stenbæk (NAV).

Arnoldo Frigessi awarded NORA's Lifetime achievement Award

Congratulations to BigInsight’s co-director Arnoldo Frigessi for being honoured with the NORA Lifetime achievement Award!

The NORA (Norwegian Artificial Intelligence Research Consortium) Lifetime achievement Award recognises an individual affiliated with a Norwegian institution who has made profound contributions to the field of Artificial Intelligence, influencing the work of many others and helping to shape the state of the art.

The NORA Award Committee said that Arnoldo Frigessi has an outstanding publication record, he is the director of Integreat, has been the director of two SFIs (including BigInsight!), and has supervised more than 40 Phd students. Arnoldo has contributed to important research in AI and has been influential in Norway and internationally through leading research centers and educating students.

As Arnoldo puts it himself: “For a better world, Statistics and Machine Learning, academics, opinions with a colour“!

Clara Bertinelli Salucci from BigInsight successfully defended her thesis

BigInsight congratulates Clara Bertinelli Salucci from BigInsight and the the Department of Mathematics, who successfully defended her thesis "Advancing data-driven diagnostics and prognostics for lithium-ion batteries: A focus on model interpretability and accuracy" for the degree of Philosophiae Doctor on April 19, 2024.

Main research findings:
Machine Learning techniques are transforming industries, yet their opaque decision-making poses challenges, especially in safety-critical applications like lithium-ion (Li-ion) batteries. Li-ion batteries are pivotal in driving sustainable transportation but are also associated with hazards, as seen in incidents such as explosions leading to particularly intense fire. Traditional testing methods for battery health are costly and disruptive. My research addresses this issue by integrating data-driven methodologies within the Battery Management Systems (BMS). These algorithms, utilising sensor data, offer an efficient monitoring of battery health and remaining life, enabling both diagnostics and prognostics. By achieving a delicate balance between interpretability and prediction accuracy, they may enhance battery performance and durability while mitigating risks associated with degradation. Moreover, the methodologies developed in this research hold promise for broader applications beyond battery diagnostics, including fields like biostatistics. By bridging the gap between interpretability and prediction accuracy, these methods offer versatile solutions for addressing complex data challenges across various domains.

Adjudication committee:

  • Professor Emeritus Bo Henry Lindquist, NTNU: Norwegian University of Science and Technology

  • Professor Sören Ehlers, Hamburg University of Technology

  • Associate Professor Thordis Thorarinsdottir, University of Oslo

 Supervisors:

  • Associate Professor Riccardo De Bin, University of Oslo

  • Professor Ingrid Kristine Glad, University of Oslo

  • Azzeddine Bakdi

  • Associate Professor Erik Vanem, University of Oslo

Ingrid Glad vinner Sverdrupprisen som "en fremragende representant for statistikkfaget"

Ingrid Glad. Foto: UiO.

Ingrid Glad, som er co-director i BigInsight, blir tildelt Sverdrupprisen som “en fremragende representant for statistikkfaget”.

Sverdrupprisen er en norsk forskningspris til minne om Erling Sverdrup. Prisen deles ut av Norsk statistisk forening. Prisen kan deles ut hvert år, men har hittil blitt delt ut under det Norske statistikermøtet, som finner sted i oddetallsår. Det deles ut to priser:

  • til en fremragende representant for statistikkfaget

  • til unge forskere (under 40 år) med beste artikkel

Hver av prisene består av et diplom og et stipend på kr 10 000.

Podkast med BigInsights Lars Henry Berge Olsen: "Fra BigInsight til Alan Turing-instituttet: En forklaring av forklaringer"

Lars Henry Berge Olsen er til vanlig PhD-student ved BigInsight og Universitetet i Oslo, og akkurat nå er han ved Alan Turing-instituttet i London. Vi snakker om hva Alan Turing-instituttet er og om Lars' egen forskning på forklarbar kunstig intelligens, som kan ligne litt på en diskusjon om hvordan en bør dele taxi-regninga.

Hør episoden i podkasten Sannsynligvis VIKTIG.

BigInsight Celebration Day was fun!

As BigInsight is approaching its final year, it was time to celebrate ourselves and our achievements. The invitational BigInsight Celebration Day took place in the beautiful Georg Sverdrups hus at the University of Oslo, on November 17, 2023. We learned from some of our successes, discussed how we have merged innovation and science, looked to the future and mingled with the BigInsight family.

Erlend Wiland-Evensen, chairman of the BigInsight board

Camilla Stoltenberg, the former head of FHI (the Norwegian Institute of Public Health) praised BigInsight’s substantial efforts with the modelling of the COVID-19 pandemic in Norway. Please see the video from Camilla Stoltenberg below.

Gunnar Rø (FHI) followed up with a talk on experiences from modelling the COVID-19 pandemic in Norway.

Idris Eckley, distinguished professor at Lancaster University and the chair of BigInsight’s Scientific Advisory Committee, also praised BigInsight’s work: «There have been fantastic contributions in terms of fundamental research, making an impact on important industrial challenges, and helping society, especially given the prominent role the people within the BigInsights team played during the pandemic.»

The two research partners NR and UiO were represented by their respective leaders, André Teigland (CEO) and Svein Stølen (Rector), who agreed were much on the importance of centers like these.

Svein Stølen, rector UiO

The strong connection between research and innovation was further demonstrated by four panels talks where an important BigInsight paper was presented, and the related innovation was discussed broadly. The talks and panels included Kjersti Aas (NR), Cathrine Pihl Lyngstad (NAV), Erlend Willand-Evensen (Gjensidige), Karl Aksel Festø (DNB), Morten Stakkeland (ABB), Erik Vanem (DNV), Martin Tveten (NR), Ingrid Glad (UiO), Alvaro Köhn-Luque (UiO) and Vessela N. Kristensen (UiO), Håvard Kvamme (Abelee). The four papers were:

  • «Explaining individual predictions when features are dependent: More accurate approximations to Shapley values». Artificial Intelligence (2021)

  • «Predicting mortgage default using convolutional neural networks». Expert Systems with Applications (2018)

  • «Scalable changepoint and anomaly detection in cross-correlated data with an application to condition monitoring». Annals of Applied Statistics (2022)

  • «Toward Personalized Computer Simulation of Breast Cancer Treatment: A Multiscale Pharmacokinetic and Pharmacodynamic Model Informed by Multitype Patient». Cancer Research (2019)

Martin Tveten

Alvaro Köhn-Luque

Morten Stakkeland

Arnoldo Frigessi, who has been an instrumental director of BigInsight until recently (he is now director of Integreat – Norwegian Centre for Knowledge-driven Machine Learning), received many kind words and gratitude for his efforts as director of BigInsight.

Arnoldo Frigessi and Anders Løland

Finally, Anders Løland (director of BigInsight) finished off by reminding ourselves that some work remains to be done in the centre in 2024, and that BigInsight’s legacy will live far beyond 2024.