Smoothed Analysis of Local Search

Local search is a powerful paradigm for finding good solutions to intractable combinatorial optimization problems. However, for many local search heuristics there exist worst-case instances on which they are extremely slow or provide solutions that are far from optimal. Smoothed analysis is a semi-random input model that has been invented to bridge the gap between poor worst-case and good empirical performance of algorithms. In smoothed analysis, an adversary picks an arbitrary input, which is then slightly randomly perturbed. In particular, smoothed analysis has been applied successfully to local search algorithms in a variety of cases. We use the 2-opt heuristic for the traveling salesman problem and the k-means method for clustering as examples to explain how local search heuristics can be analyzed in the framework of smoothed analysis. For both algorithm, as for many other local search algorithms, the worst-case running time is exponential in the input size, but polynomial in the framework of smoothed analysis.

[1]  J. Yukich.  Probability theory of classical Euclidean optimization problems , 1998 .

[2]  Leena Jain,et al.  Traveling Salesman Problem: A Case Study , 2012, BIOINFORMATICS 2012.

[3]  Hiroshi Imai,et al.  Variance-based K-clustering algorithms by Voronoi Diagrams and randomization , 2000 .

[4]  Wansoo T. Rhee.  A Matching Problem and Subadditive Euclidean Functionals , 1993 .

[5]  Yuval Peres,et al.  Local max-cut in smoothed polynomial time , 2017, STOC.

[6]  Berthold Vöcking,et al.  Smoothed Analysis of the 2-Opt Algorithm for the General TSP , 2016, ACM Trans. Algorithms.

[7]  Bodo Manthey,et al.  Worst-Case and Smoothed Analysis of k-Means Clustering with Bregman Divergences , 2009, J. Comput. Geom..

[8]  Craig A. Tovey,et al.  New Results on the Old k-opt Algorithm for the Traveling Salesman Problem , 1999, SIAM J. Comput..

[9]  Shang-Hua Teng,et al.  Smoothed analysis of algorithms: why the simplex algorithm usually takes polynomial time , 2001, STOC '01.

[10]  Sergei Vassilvitskii,et al.  Worst-case and Smoothed Analysis of the ICP Algorithm, with an Application to the k-means Method , 2006, 2006 47th Annual IEEE Symposium on Foundations of Computer Science (FOCS'06).

[11]  ATSPDavid S. JohnsonAT.  Experimental Analysis of Heuristics for the Stsp , 2001 .

[12]  Bodo Manthey,et al.  Smoothed Analysis of the 2-Opt Heuristic for the TSP: Polynomial Bounds for Gaussian Noise , 2013, ISAAC.

[13]  Berthold Vöcking,et al.  Worst Case and Probabilistic Analysis of the 2-Opt Algorithm for the TSP , 2007, SODA '07.

[14]  Bodo Manthey,et al.  Smoothed Analysis of the k-Means Method , 2011, JACM.

[15]  Bodo Manthey,et al.  Towards Understanding the Smoothed Approximation Ratio of the 2-Opt Heuristic , 2015, ICALP.

[16]  Andrea Vattani,et al.  k-means Requires Exponentially Many Iterations Even in the Plane , 2008, SCG '09.

[17]  Heiko Röglin,et al.  Smoothed Analysis of Local Search for the Maximum-Cut Problem , 2017, ACM Trans. Algorithms.

[18]  Mihalis Yannakakis,et al.  Simple Local Search Problems That are Hard to Solve , 1991, SIAM J. Comput..

[19]  Robert Elsässer,et al.  Settling the Complexity of Local Max-Cut (Almost) Completely , 2010, ICALP.