Research and Projects

under renovation

Density-Based Mapper for Temporal Topic Modeling:

Given a dataset, the Mapper algorithm allows one to compute a graph that describes the topology of the dataset, in terms of a specific lens function .

For a dataset with time-information, using time as the lens function for Mapper produces graphs that describe the evolution of topics in the dataset over time. Unfortunately, for real datasets with variable local density, Mapper struggles to produce robust graphs.

This project demonstrates how to improve the robustness of Mapper by generalizing the algorithm, incorporating local density. We call our proposed generalization density-based Mapper , and provide a reference implementation.

Notes:

- Mirror Symmetry for toric varieties and other GIT quotients.

Mentorship: