Master Thesis for the Master Program Computational Science from the University of Vienna
Title - Data Driven Modeling using Explicit Interpolation and Extrapolation Behaviour
Overview
I am trying to incorporate constraints on monotonicity and shape (e.g. convexity) into the learning process to control interpolation and extrapolation behaviour. The focus should be on local methods, utilizing basis functions (e.g. LOLIMOT, Splines,...)
Algorithms
- Interpolated Look-up Tables - TensorFlow Lattice
- Generalized Additive Models - pyGAM
- Local Linear Models - LOLIMOT
- Symbolic neural networks - Equation Learner +
- Gaussian Process Regression