Approximation and Online Algorithms with Applications in Computational Biology and Computational Geometry

Abstract: The main contributions of this thesis are in the area of approximation and online algorithm design and derivation of lower bounds on the approximability for a number of combinatorial optimization problems with applications in computational biology and computational geometry. Approximation and online algorithms are fundamental tools used to deal with computationally hard problems and problems in which the input is gradually disclosed over time. The thesis is divided into two parts. In the first part we study some problems where one seeks to find a (possibly hierarchical) clustering of a set of objects such that particular objective functions are either maximized or minimized. We study the computational complexity, present polynomial-time approximation algorithms or derive bounds on the allowed degree of approximability achievable in polynomial time for these problems. The second part is devoted to decision making based on only partial information on the input. We present online algorithms for two problems with applications in the area of robotics and use the well established approach of competitive analysis to analyze their performance. We also study lower bounds, aiming at establishing to what extent the problems can be approximated.