Algorithms for stable and perturbation-resilient problems

@article{Angelidakis2017AlgorithmsFS,
  title={Algorithms for stable and perturbation-resilient problems},
  author={Haris Angelidakis and Konstantin Makarychev and Yury Makarychev},
  journal={Proceedings of the 49th Annual ACM SIGACT Symposium on Theory of Computing},
  year={2017}
}
We study the notion of stability and perturbation resilience introduced by Bilu and Linial (2010) and Awasthi, Blum, and Sheffet (2012). A combinatorial optimization problem is α-stable or α-perturbation-resilient if the optimal solution does not change when we perturb all parameters of the problem by a factor of at most α. In this paper, we give improved algorithms for stable instances of various clustering and combinatorial optimization problems. We also prove several hardness results. We… 

Perturbation Resilient Clustering for k-Center and Related Problems via LP Relaxations

This model shows that the unified MST and dynamic programming based algorithm proposed by Angelidakis et.

k-center Clustering under Perturbation Resilience

This work provides strong positive results both for the asymmetric and symmetric k-center problems under a natural input stability (promise) condition called α-perturbation resilience and provides algorithms that give strong guarantees simultaneously for stable and non-stable instances.

Clustering Perturbation Resilient Instances

This work considers stable instances of Euclidean $k-means that have provable polynomial time algorithms for recovering optimal cluster and proposes simple algorithms with running time linear in the number of points and the dimension that provably recover the optimal clustering.

Exact Algorithms and Lower Bounds for Stable Instances of Euclidean k-Means

It is shown that for any fixed \epsilon>0, (1+\ep silon)-stable instances of k-Means in doubling metrics can be solved in polynomial time and under a plausible PCP hypothesis this is essentially tight.

Certified Algorithms: Worst-Case Analysis and Beyond

The notion of a certified algorithm is introduced, its properties are described, a framework for designing certified algorithms is presented, and examples of certified algorithms for Max Cut/Min Uncut, Minimum Multiway Cut, k-medians and k-means are provided.

On the Local Structure of Stable Clustering Instances

It is obtained that the widely-used Local Search algorithm has strong performance guarantees for both the tasks of recovering the underlying optimal clustering and obtaining a clustering of small cost.

On Perturbation Resilience of Non-uniform k-Center

The NUkC problem under perturbation resilience, which was introduced by Bilu and Linial, is shown to be polynomial time solvable when the r_i$'s belong to a constant sized set, however, it is shown that perturbations resilience does not help in the general case.

Bilu-Linial stability, certified algorithms and the Independent Set problem

A general result is proved showing that the integrality gap of convex relaxations of several maximization problems reduces dramatically on stable instances of Node Multiway Cut.

On Euclidean k-Means Clustering with α-Center Proximity

It is shown that for any α′ > 1, there exists an α 6 α′, (α > 1), and an ε0 > 0 such that minimizing the k-means objective over clusterings that satisfy α-center proximity is NP-hard to approximate within a multiplicative (1 +ε0) factor.

Stability and Recovery for Independence Systems

This work considers perturbation-stable instances, in the sense of Bilu and Linial, and precisely identifies the stability threshold beyond which these algorithms are guaranteed to recover the optimal solution, and resolves the worst-case approximation guarantee of local search in p-extendible systems.

References

SHOWING 1-10 OF 22 REFERENCES

k-center Clustering under Perturbation Resilience

This work provides strong positive results both for the asymmetric and symmetric k-center problems under a natural input stability (promise) condition called α-perturbation resilience and provides algorithms that give strong guarantees simultaneously for stable and non-stable instances.

Data stability in clustering: A closer look

Center-based clustering under perturbation stability

Bilu-Linial Stable Instances of Max Cut and Minimum Multiway Cut

It is proved that there is no robust polynomial-time algorithm for γ-stable instances of Max Cut when γ < α SC(n/2), where αSC is the best approximation factor for Sparsest Cut with non-uniform demands, and it is shown that the standard SDP relaxation for Max Cut is integral if [EQUATION].

On the Complexity of the Metric TSP under Stability Considerations

It is shown that for γ ≥ 1.8 a simple greedy algorithm computes the optimum Hamiltonian tour for every γ-stable instance of the Δ-TSP, whereas a simple local search algorithm can fail to find the optimum even if γ is arbitrary.

CS264: Beyond Worst-Case Analysis Lecture #8: Exact Recovery in Stable Cut Instances

This lecture makes explicit a theme introduced surreptitiously last lecture: exact recovery, to explain when NP -hard problems can be solved exactly via linear programming.

Multiway cut, pairwise realizable distributions, and descending thresholds

New approximation algorithms for the Multiway Cut problem are designed, improving the previously known factor of 1.32388 by introducing a new rounding scheme: (3) single-coordinate cuts with descending thresholds, and design an algorithm that achieves a factor of [EQUATION].

On the practically interesting instances of MAXCUT

This work investigates practically interesting instances of MAXCUT, viewed as a clustering problem, and shows how to solve in polynomial time distinguished, metric, expanding and dense instances ofMAXCUT under mild stability assumptions.

Linear degree extractors and the inapproximability of max clique and chromatic number

New extractors which require only log n + O(1) additional random bits for sources with constant entropy rate are constructed, and dispersers, which are similar to one-sided extractors, are built, which use an arbitrarily small constant times log n additional Randomness Extractor to within n1-ε are NP-hard.

Robust algorithms for restricted domains