Non-Euclidean Machine Learning

Inspired by the work of Sang Eon Park et al. (2208.05484), I developed methods for hyperbolic machine learning in jet physics analysis. This work has developed from solely hyperbolic geometries to mixed-curvature machine learning employing advanced methods in geometric learning while utilizing manifold transformers.

I recently presented some preliminary results at the 2023 ML4Jets conference at DESY (click here to view the presentation):

Particle jets exhibit tree-like structures through stochastic showering and hadronization. The hierarchical nature of these structures aligns naturally with hyperbolic space, a non-Euclidean geometry that captures hierarchy intrinsically. Drawing upon the foundations of geometric learning, we introduce hyperbolic transformer models tailored for tasks relevant to jet analyses, such as classification and representation learning. Through jet embeddings and jet tagging evaluations, our hyperbolic approach outperforms its Euclidean counterparts. These findings underscore the potential of using hyperbolic geometric representations in advancing jet physics analyses.

Paper in progress…