핵심 개념
Expected complexity of persistent homology computation via matrix reduction is analyzed for various filtrations, showing sparser matrices than worst-case predictions.
초록
The article studies algorithmic complexity in computing persistent homology for different filtrations.
Upper bounds for fill-in of boundary matrices are proven after matrix reduction.
Reduction algorithm performance is better than worst-case predictions in realistic datasets.
Expected fill-in and cost of matrix reduction are analyzed for ˇCech, Vietoris–Rips, and Erd˝os–R´enyi filtrations.
Results show improved performance compared to worst-case scenarios.
Concrete examples and experiments support the theoretical analysis.
통계
경계 행렬의 비어 있지 않은 항의 수에 대한 상한을 증명합니다.
ˇCech, Vietoris–Rips 및 Erd˝os–R´enyi 필터링 후 축소된 행렬의 평균 채움 수에 대한 예상치를 제시합니다.
행렬 축소의 비용에 대한 한계를 제시합니다.
인용구
"Our bounds show that the reduced matrix is expected to be significantly sparser than what the general worst-case predicts."
"The reduction algorithm scales closer to linear in practice, leading to the hypothesis that the worst-case examples are somewhat pathological."