'Topological data analysis in machine learning'
Trial lecture in connection with recruitment for ternure track position in machine learning:
Hovedinnhold
Abstract:
The goal of topological data analysis is to find structure in data with a focuson topological features such as shape and connectivity. Persistent homology is a method to quantify topological features using a multiscale approach. This has the advantage that in addition to finding features in data, persistent homology also finds the scales at wich these features occur. On the other hand, this multiscale approach can be computationally expensive.
In this talk we will give an introduccion to persistent homology, show how approximation algorithms can be used to reduce the computational cost and how topological data analysis can be used together with other machine learning algorithms for clustering, dimensionality reduction and clasification. Finally we will highlight some open questions
