Project
For many optimization problems that arise in logistics, information retrieval, and other contexts the classical theory of algorithms has lost its grip on reality because it is based on a pessimistic worstcase perspective, in which the performance of an algorithm is solely measured by its behavior on the worst possible input. This does not take into consideration that worstcase inputs are often rather contrived and occur only rarely in practical applications. It led to the situation that for many problems the classical theory is not able to differentiate meaningfully between different algorithms. Even worse, for some important problems it recommends algorithms that perform badly in practice over algorithms that work well in practice only because the artificial worstcase performance of the latter ones is bad.
We will study classic optimization problems (traveling salesperson problem, linear programming, etc.) as well as problems coming from machine learning and information retrieval. All these problems have in common that the practically most successful algorithms have a devastating worstcase performance even though they clearly outperform the theoretically best algorithms.
Only in recent years a paradigm shift towards a more realistic and robust algorithmic theory has been initiated. In this project we will develop and explore novel theoretical approaches to reconcile theory and practice. We will, in particular, analyze algorithms and optimization problems in the framework of smoothed analysis. In this model, inputs are generated in two steps: first, an adversary chooses an arbitrary instance, and then this instance is slightly perturbed at random. The smoothed performance of an algorithm is defined to be the worst expected performance the adversary can achieve. This model can be viewed as a less pessimistic worstcase analysis, in which the randomness rules out pathological worstcase instances that are rarely observed in practice but dominate the worstcase analysis.
Team Members
 Heiko Röglin (Principal Investigator)
 Michael Etscheid (PhD student)
 Carsten Fischer (PhD student)
 Matthias Mnich (Postdoc)
Publications
[13] 
The Alternating Stock Size Problem and the Gasoline Puzzle
To appear in Proc. of the 24th ESA (Aarhus, Denmark), 2016.


[12] 
The Smoothed Number of Paretooptimal Solutions in Noninteger Bicriteria Optimization
To appear in Proc. of the 13th TAMC (Xi'an, China), 2016.


[11] 
Probabilistic Analysis of the Dual NextFit Algorithm for Bin Covering
In Proc. of the 12th LATIN (Ensenada, Mexico), pp. 469482, 2016.


[10] 
New Deterministic Algorithms for Solving Parity Games
In Proc. of the 12th LATIN (Ensenada, Mexico), pp. 634645, 2016.


[9] 
Improved Analysis of CompleteLinkage Clustering
In Proc. of the 23rd ESA (Patras, Greece), pp. 656667, 2015.


[8] 
Smoothed Analysis of the Squared Euclidean MaximumCut Problem
In Proc. of the 23rd ESA (Patras, Greece), pp. 509520, 2015.


[7] 
Polynomial Kernels for Weighted Problems
In Proc. of the 40th MFCS (Milano, Italy), pp. 287298, 2015.


[6] 
Solving Totally Unimodular LPs with the Shadow Vertex Algorithm
In Proc. of the 32nd STACS (Munich, Germany), pp. 171183, 2015.


[5] 
Smoothed Analysis of Local Search for the MaximumCut Problem
In Proc. of the 25th SODA (Portland, USA), pp. 882889, 2014.


[4] 
Performance Guarantees for Scheduling Algorithms under Perturbed Machine Speeds
In Proc. of the 24th ISAAC (Hong Kong), pp. 207217, 2013.
Discrete Applied Mathematics, 2014.


[3] 
Finding Short Paths on Polytopes by the Shadow Vertex Algorithm
In Proc. of the 40th ICALP (Riga, Latvia), pp. 279290, 2013.


[2] 
Smoothed Analysis of Belief Propagation for MinimumCost Flow and Matching
In Proc. of the 7th WALCOM (Kharagpur, India), pp. 182193, 2013.
Journal of Graph Algorithms and Applications, 17(6):647670, 2013.


[1] 
Smoothed Analysis of the Successive Shortest Path Algorithm
In Proc. of the 24th SODA (New Orleans, USA), pp. 11801189, 2013.
