Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a multi-scale metric \( d_\rho \) for strings that aggregates angle distances between \( n \)-gram count vectors using exponential weights. The authors benchmark \( d_\rho \) against traditional edit and \( n \)-gram baselines in DBSCAN clustering, demonstrating superior performance and robustness, particularly under tandem-repeat stutters. Key results include a linear-time suffix-tree algorithm for efficient evaluation and a comprehensive analysis of the metric's stability properties and isometries.
A novel string metric outperforms traditional baselines in clustering while maintaining robustness against common perturbations.
We define a multi-scale metric $d_ρ$ on strings by aggregating angle distances between all $n$-gram count vectors with exponential weights $ρ^n$. We benchmark $d_ρ$ in DBSCAN clustering against edit and $n$-gram baselines, give a linear-time suffix-tree algorithm for evaluation, prove metric and stability properties (including robustness under tandem-repeat stutters), and characterize isometries.