Core Concepts
The author explores the complexity of Earth Mover's Distance under translation, presenting algorithms and lower bounds for different metrics and dimensions.
Abstract
The content delves into the Earth Mover's Distance under Translation (EMDuT), discussing its significance, algorithms, conditional lower bounds, and applications. The study focuses on fine-grained complexity in various dimensions and metrics, providing insights into optimal translations and approximation algorithms.
Key points include:
Introduction to Earth Mover's Distance (EMD) as a similarity measure.
EMDuT as a translation-invariant version minimizing distance over translations.
Algorithms presented for EMDuT in different dimensions and metrics.
Conditional lower bounds based on Orthogonal Vectors Hypothesis (OVH) and Exponential Time Hypothesis (ETH).
Open problems related to fast approximation algorithms for EMDuT in higher dimensions.
The study showcases the importance of translation-invariant distance measures like EMDuT in geometric algorithms research.
Stats
For EMDuT in R1, an eO(n2)-time algorithm is presented.
For EMDuT in Rd with L1 and L∞ metric, an eO(n2d+2)-time algorithm is provided.
Quotes
"No algorithm solves the Orthogonal Vectors problem in time O(n2−δdc) for any constants δ, c > 0." - Orthogonal Vectors Hypothesis