A Topological Machine Learning Pipeline for Classification

In this work, we develop a pipeline that associates Persistence Diagrams to digital data via the most appropriate filtration for type of considered. Using grid search approach, determines optimal representation methods and parameters. The development such topological Machine Learning involves two crucial steps strongly affect its performance: firstly, must be represented as an algebraic object with proper associated in order compute summary, Diagram. Secondly, persistence diagram transformed suitable introduced algorithm. We assess performance our pipeline, parallel, compare different on popular benchmark datasets. This work is first step toward both easy ready-to-use classification using persistent homology Learning, understand theoretical reasons why, given dataset task performed, pair (filtration, representation) better than another.