Home
Berent Ånund Strømnes Lunde's picture

Berent Ånund Strømnes Lunde

Associate Professor
  • E-mailBerent.Lunde@uib.no
  • Visitor Address
    Realfagbygget, Allégaten 41
  • Postal Address
    Postboks 7803
    5020 Bergen

Machine learning / Information theory / Computational statistics

I develop information theory for algorithms in machine learning and computational statistics. My conjecture is that, through a deeper understanding of the mathematical and statistical properties of ML-algorithms, it is possible to device smarter and more data- and information-adaptive ML-algorithms. Currently I work on theory and methods to avoid all types of manual tuning in gradient tree boosting-type methods.

Mathmatical finance / Actuarial mathematics

I seek more extensive usage of machine learning and advanced statistical modelling in the applied actuarial field. I believe a competetive market will require the industry to capitalize on modern statistical methodology, and see methods work in symbiosis on both risk-assessments, customer behaviour and more, to optimize value for stakeholders. To this end, the methods needs to be safe and understandable for practitioners to apply, and robustly implemented for production environments.

Lecture
  • Show author(s) 2018. Boosting i forsikring.
Academic lecture
  • Show author(s) 2019. Information criteria for gradient boosted trees: Adaptive tree size and early stopping.
  • Show author(s) 2019. An information criterion for gradient boosted trees.
  • Show author(s) 2019. An information criterion for gradient boosted trees.
  • Show author(s) 2018. Saddlepoint adjusted inversion of characteristic functions.
  • Show author(s) 2018. Information efficient gradient tree boosting.
  • Show author(s) 2018. Information efficient gradient tree boosting.
  • Show author(s) 2018. Finance in the frequency domain.
  • Show author(s) 2017. Likelihood Estimation of Jump-Diffusions: Extensions from Diffusions to Jump-Diffusions, Implementation with Automatic Differentiation, and Applications.
Masters thesis
  • Show author(s) 2016. Likelihood Estimation of Jump-Diffusions: Extensions from Diffusions to Jump-Diffusions, Implementation with Automatic Differentiation, and Applications.

More information in national current research information system (CRIStin)