Skip to content

Learned Filtration Metrics for TDA

Home / Open Questions / Learned Filtration Metrics for TDA

Background: Topological data analysis (TDA) typically relies on pre-defined filtration metrics (like Pearson distance) to characterize the structure of data populations, which may not capture complex non-linear dependencies.

Question / Future Work: Developing a learnable filtration metric for TDA could allow for more robust recovery of latent topological structures in time-series populations, potentially overcoming limitations imposed by standard distance metrics in forecasting contexts.

Why It Matters: Moving beyond fixed distance metrics is a significant hurdle in making TDA more effective for high-dimensional, non-linear forecasting data.

Metadata & Links