TDA crash course: theory and practice for ML applications
09:00-16:30 January 26

Full day
Intermediate level

Details

We will introduce elements of the theory of topological data analysis (TDA) and show the participants how to apply standard topological data pipelines, such as persistent homology and Mapper, to datasets of different origin (spatial data, word embeddings, brain networks). We will also show how TDA can be used for feature engineering or as a preprocessing step in standard machine learning pipelines, and how to assess the significance of the unearthed features. Finally, since TDA outputs can sometimes be of difficult interpretation, we will guide the participants through some common pitfalls in the interpretation and provide some examples of successful applications.

The workshop will consist of: an introductory section on TDA methods; hands-on exercises focused on implementing the required pipelines and performing the analysis; a session on validation methods and null models for TDA; a session on strengths and weaknesses of topological methods.

We will use Jupyter notebooks in Python for analysis and visualization.


Outcome

  • Participants will be able to set up a TDA pipeline
  • Participants will be able to interpret the relevance of the results based on both theory and practice
  • Participants will have developed ready-to-use code, which they will be able to adapt to similar applications

Prerequisites

  • basic Python programming skills (laptop with Python 2.7/3 installed)
  • basic Jupyter notebook skills
  • basic git/data management skills
  • notions of linear algebra and elementary topology are helpful but not needed

Organizers

Giovanni Petri

Research Leader, ISI Foundation

Twitter  ยท  Website