Lecture plan MA8701V2023

This plan is tentative and may change!

Part 1: Core concepts (3 weeks: 09.01, 13.01, 16.01, 20.01, 23.01, 27.01)

  • L1: Overview of course, getting to know each other (much compulsory work in groups), review on linear regression, decision theoretic framework for regression
  • L2: Review of classification, decision theoretic framework for classification, bias-variance trace-off
  • W2 (L3-L4): Model selection and model assessment results
  • W3 (L5-L6): Missing data: mechanisms of missingness, single imputation methods including Bayesian linear regression, predictive mean matching, Rubins rules, fully conditional specification of imputation models.

Part 2: Shrinkage and regularization in LM and GLM (3 weeks: 30.01, 03.02, 06.02, 10.02, 13.02, 17.02,20.02)

  • L7: Ridge for linear models
  • L8: Lasso for linear models
  • L9: Variants of lasso and ridge
  • L10: Ridge and lasso for logistic regression (and GLM)
  • L11+L12: Statistical inference for lasso/ridge and selective inference

Part 3: Ensembles (4 weeks: 24.02, 27.02, 03.03, 06.03, 10.03, 13.03, no lecture 17.03 )

  • 24.02: L13 Wisdom of the crowds, bagging, trees, random forest
  • 27.02: L14 Adaboost and gradient tree boosting
  • Watch video 03.03 by yourself (so no lecture) - link to talk by Berent Lunde in MA8702021V on Xgboost will be sent to participants
  • 06.03: L15 Stacked ensembles
  • 10.03: L16 Hyperparameter tuning
  • 13.03: L17 Uncertainty assessment

Part 4: Explainable AI/Interpretable Machine Learning General introduction to the topic, partial dependence plots, ALE plots, LIME, Shapley values and counterfactuals.

  • 20.03: L18 digitally on zoom with Kjersti from Oslo
  • 24.03: L19 digitally on zoom with Kjersti from Oslo
  • 27.03 (10.15-12) L20 physically in S21 with Kjersti in Trondheim

No lecture 31.03 - happy easter!

Part 5: Closing

  • Article presentations: 14.04, 17.04, 21.04
  • Summing up lecture: 24.04
2023-03-24, Mette Langaas