Bayesian Isotonic Calibration and its Optimisation
Name
Karl Riis
Abstract
The work focused on describing and optimising the Bayesian isotonic calibration algorithm created by Mari-Liis Allikivi and Meelis Kull. The algorithm needed optimisation because it was slow on large datasets. Diļ¬erent mathematical concepts related to the algorithm are described in detail. The algorithm is then improved with various techniques which make it more stable and faster. Finally the algorithm is compared to isotonic calibration and logistic regression using synthetic data.
Graduation Thesis language
Estonian
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Meelis Kull
Defence year
2019