DataScience@BI seminar with Ingrid Dæhlen
Doctoral Research Fellow Ingrid Dæhlen, University of Oslo, focus on theoretical machine learning. She will give a research talk titled "The asymptotic effect of tuning parameters"
- Starts:12:00, 8 October 2024
- Ends:13:00, 8 October 2024
- Location:BI - campus Oslo, room: B3 inner area - next to meeting room B3i-108 or Zoom
- Contact:Siri Johnsen (siri.johnsen@bi.no)
"The asymptotic effect of tuning parameters"
DataScience@BI invites Doctoral Research Fellow Ingrid Dæhlen, University of Oslo, to give a research talk titled "The asymptotic effect of tuning parameters".
Abstract:
Tuning parameters are parameters involved in an estimating procedure for the purpose of reduce the risk of some other estimator. Examples include the degree of penalization in penalized regression and likelihood problems, as well as the balance parameter in hybrid methods. Typically tuning parameters are set to the minimizers of some estimator of the risk, a step which introduces additional randomness and makes standard methodology inapplicable. We derive precise asymptotic theory for this situation. Our framework allows for smooth, but otherwise arbitrary, loss functions and for the risk to be estimated by cross validation procedures. Results include consistency of the optimal estimator towards a well-defined quantity and asymptotic normality after proper scaling and centring. We give explicit forms and estimators for the limiting variance matrix and results sharply characterizing the distance from the training error to the cross-validated estimator of the risk.
Key research areas: Theoretical machine learning.