TBA

The path signature is a powerful tool for solving regression problems on path space, i.e., for computing conditional expectations $\mathbb{E}[Y | X]$ when the random variable $X$ is a stochastic process -- or a time-series. We provide new theoretical convergence guarantees for two different, complementary approaches to regression using signature methods. In the context of global regression, we show that linear functionals of the robust signature are universal in the $L^p$ sense in a wide class of examples. In addition, we present a local regression method based on signature semi-metrics, and show universality as well as rates of convergence. 

Based on joint works with Davit Gogolashvili, Luca Pelizzari, and John Schoenmakers.

Please note: The MCF seminar usually takes place on Thursdays from 16:00 to 17:00 in L5. However, for this week, the timing will be changed to 15:00 to 16:00.

Merge trees are a topological descriptor of a filtered space that enriches the degree zero barcode with its merge structure. The space of merge trees comes equipped with an interleaving distance d_I , which prompts a naive question: is the interleaving distance between two merge trees equal to the bottleneck distance between their corresponding barcodes? As the map from merge trees to barcodes is not injective, the answer as posed is no, but as proposed by Gasparovic et al., we explore intrinsic metrics d_I and d_B realized by infinitesimal path length in merge tree space, which do indeed coincide. This result suggests that in some special cases the bottleneck distance (which can be computed quickly) can be substituted for the interleaving distance (in general, NP-hard).