Learnability of Timescale Graphical Event Models

TL;DR

This paper investigates the learnability of Timescale Graphical Event Models by proposing heuristics for hyper-parameter selection, refining a distance measure, and benchmarking their effectiveness on synthetic data.

Contribution

It introduces new heuristics for hyper-parameter tuning, refines an existing distance measure, and provides a comprehensive benchmark for Timescale Graphical Event Models.

Findings

Heuristics improve structure learning accuracy

Refined distance measure enhances model evaluation

Benchmark results guide applicability of methods

Abstract

This technical report tries to fill a gap in current literature on Timescale Graphical Event Models. I propose and evaluate different heuristics to determine hyper-parameters during the structure learning algorithm and refine an existing distance measure. A comprehensive benchmark on synthetic data will be conducted allowing conclusions about the applicability of the different heuristics.