Performance Mining

Bachelor & Master Thesis

“On which part of the dataset do (foundation) models work best?”

What if we trained on a subset of our data? Would our model perform better? Do foundation models like TabPFN have similar characteristics?

To answer this, we need efficient exceptional model mining algorithms that support some kind of transfer learning. This is a thesis combining many exciting concepts from state-of-the-art machine learning, deep learning, foundation models, and exceptional model mining.