Ran Hu,Divy H. Kanani,Jingru Zhang
Abstract:In this paper, we consider the (weighted) one-center problem of uncertain points on a cactus graph. Given are a cactus graph $G$ and a set of $n$ uncertain points. Each uncertain point has $m$ possible locations on $G$ with probabilities and a non-negative weight. The (weighted) one-center problem aims to compute a point (the center) $x^*$ on $G$ to minimize the maximum (weighted) expected distance from $x^*$ to all uncertain points. No previous algorithm is known for this problem. In this paper, we propose an $O(|G| + mn\log mn)$-time algorithm for solving it. Since the input is $O(|G|+mn)$, our algorithm is almost optimal.
What problem does this paper attempt to address?
### What problem does this paper attempt to solve?
This paper mainly studies the weighted one - center problem of uncertain points on cactus graphs. Specifically, given a cactus graph \(G\) and a set of \(n\) uncertain points, each uncertain point has \(m\) possible positions, and each position has a probability and a non - negative weight. The goal of the weighted one - center problem is to find a point \(x^*\) on graph \(G\) such that the maximum weighted expected distance from \(x^*\) to all uncertain points is minimized.
#### Problem background
1. **Uncertain data**: Many real - world measurement data are accompanied by uncertainty. For example, in the facility location problem, the demand points may be uncertain.
2. **Cactus graph**: A cactus graph is a special type of graph in which any two cycles do not share edges. This type of graph is relatively common in practical applications such as facility location.
3. **One - center problem**: For deterministic points, the one - center problem is to find a point such that the maximum distance from it to all other points is minimized. In the case of uncertain points, the goal is to minimize the maximum weighted expected distance.
#### Research motivation
- On tree - shaped networks, there are already linear - time algorithms for the weighted one - center problem.
- For uncertain points on cactus graphs, there were no known algorithms to solve this problem before.
#### Main contributions
- Proposed an algorithm with a time complexity of \(O(|G|+mn\log mn)\) to solve the weighted one - center problem on cactus graphs, where \(|G|\) is the size of graph \(G\), \(m\) is the number of possible positions of each uncertain point, and \(n\) is the number of uncertain points.
- This algorithm is almost optimal because the input scale is \(O(|G| + mn)\).
### Formula summary
1. **Weighted expected distance**:
\[
Ed(P_i,x)=w_i\cdot\sum_{j = 1}^{m}f_{ij}d(p_{ij},x)
\]
where \(w_i\) is the weight of point \(P_i\), \(f_{ij}\) is the probability that \(P_i\) appears at position \(p_{ij}\), and \(d(p_{ij},x)\) is the distance from position \(p_{ij}\) to point \(x\).
2. **Objective function**:
\[
\text{minimize }\max_{1\leq i\leq n}Ed(P_i,x)
\]
### Method overview
1. **Vertex - constrained instance**: Through Lemma 5, a general instance can be transformed into a vertex - constrained instance in linear time, that is, all positions are on the vertices of the graph.
2. **Binary search**: Utilize the structural characteristics of cactus graphs (such as block graphs and skeleton trees) to determine the block containing the center point through binary search.
3. **Dynamic convex hull data structure**: When dealing with the one - center problem on cycles, use a dynamic convex hull data structure to efficiently calculate the local center point.
### Conclusion
This paper proposes an efficient algorithm to solve the weighted one - center problem of uncertain points on cactus graphs, fills this gap in the field, and provides important theoretical and practical progress.