Michael Elberfeld

Algorithmic Meta Theorems for Circuit Classes of Constant and Logarithmic Depth

An algorithmic meta theorem for a logic and a class C of structures states that all problems expressible in this logic can be solved eciently for inputs from C. The prime example is Courcelle’s Theorem, which states that monadic second-order (mso) definable problems are linear-time solvable on graphs of bounded tree width. We contribute new algorithmic meta theorems, which state that mso-definable problems are (a) solvable by uniform constant-depth circuit families (AC 0 for decision problems and TC 0 for counting problems) when restricted to input structures of bounded tree depth and (b) solvable by uniform logarithmic-depth circuit families (NC 1 for decision problems and #NC 1 for counting problems) when a tree decomposition of bounded width in term representation is part of the input. Applications of our theorems include a TC 0 -completeness proof for the unary version of integer linear programming with a fixed number of equations and extensions of a recent result that counting the number of accepting paths of a visible pushdown automaton lies in #NC 1 . Our main technical contributions are a new tree automata model for unordered, unranked, labeled trees; a method for representing the tree automata’s computations algebraically using convolution circuits; and a lemma on computing balanced width-3 tree decompositions of trees in TC 0 , which encapsulates most of the technical diculties surrounding earlier results connecting tree automata and NC 1 . 1998 ACM Subject Classification F.1.3 Complexity Measures and Classes

Optimally orienting physical networks

In a network orientation problem, one is given a mixed graph, consisting of directed and undirected edges, and a set of source-target vertex pairs. The goal is to orient the undirected edges so that a maximum number of pairs admit a directed path from the source to the target. This NP-complete problem arises in the context of analyzing physical networks of protein-protein and protein-DNA interactions. While the latter are naturally directed from a transcription factor to a gene, the direction of signal flow in protein-protein interactions is often unknown or cannot be measured en masse. One then tries to infer this information by using causality data on pairs of genes such that the perturbation of one gene changes the expression level of the other gene. Here we provide a first polynomial-size ILP formulation for this problem, which can be efficiently solved on current networks. We apply our algorithm to orient protein-protein interactions in yeast and measure our performance using edges with known orientations. We find that our algorithm achieves high accuracy and coverage in the orientation, outperforming simplified algorithmic variants that do not use information on edge directions. The obtained orientations can lead to a better understanding of the structure and function of the network.

Efficient Algorithms for String-Based Negative Selection

String-based negative selection is an immune-inspired classification scheme: Given a self-set S of strings, generate a set D of detectors that do not match any element of S . Then, use these detectors to partition a monitor set M into self and non-self elements. Implementations of this scheme are often impractical because they need exponential time in the size of S to construct D . Here, we consider r -chunk and r -contiguous detectors, two common implementations that suffer from this problem, and show that compressed representations of D are constructible in polynomial time for any given S and r . Since these representations can themselves be used to classify the elements in M , the worst-case running time of r -chunk and r -contiguous detector based negative selection is reduced from exponential to polynomial.

On the space complexity of parameterized problems

Parameterized complexity theory measures the complexity of computational problems predominantly in terms of their parameterized time complexity. The purpose of the present paper is to demonstrate that the study of parameterized space complexity can give new insights into the complexity of well-studied parameterized problems like the feedback vertex set problem. We show that the undirected and the directed feedback vertex set problems have different parameterized space complexities, unless L = NL; which explains why the two problem variants seem to necessitate different algorithmic approaches even though their parameterized time complexity is the same. For a number of further natural parameterized problems, including the longest common subsequence problem and the acceptance problem for multi-head automata, we show that they lie in or are complete for different parameterized space classes; which explains why previous attempts at proving completeness of these problems for parameterized time classes have failed.

Approximation algorithms for orienting mixed graphs

Graph orientation is a fundamental problem in graph theory that has recently arisen in the study of signaling-regulatory pathways in protein networks. Given a graph and a list of ordered source-target vertex pairs, it calls for assigning directions to the edges of the graph so as to maximize the number of pairs that admit a directed source-to-target path. When the input graph is undirected, a sub-logarithmic approximation is known for the problem. However, the approximability of the biologically-relevant variant, in which the input graph has both directed and undirected edges, was left open. Here we give the first approximation algorithm to this problem. Our algorithm provides a sub-linear guarantee in the general case, and logarithmic guarantees for structured instances.

On the Approximability of Reachability-Preserving Network Orientations

Abstract We introduce a graph-orientation problem arising in the study of biological networks. Given an undirected graph and a list of ordered source–target vertex pairs, the goal is to orient the graph such that a maximum number of pairs admit a directed source-to-target path. We study the complexity and approximability of this problem. We show that the problem is -hard even on star graphs and hard to approximate to within some constant factor. On the positive side, we provide an Ω(log log n/log n) factor approximation algorithm for the problem on n-vertex graphs. We further show that for any instance of the problem there exists an orientation of the input graph that satisfies a logarithmic fraction of all pairs and that this bound is tight up to a constant factor. Our techniques also lead to constant-factor approximation algorithms for some restricted variants of the problem.

Computational Complexity of Perfect-Phylogeny-Related Haplotyping Problems

Haplotyping, also known as haplotype phase prediction, is the problem of predicting likely haplotypes based on genotype data. This problem, which has strong practical applications, can be approached using both statistical as well as combinatorial methods. While the most direct combinatorial approach, maximum parsimony, leads to NP-complete problems, the perfect phylogeny model proposed by Gusfield yields a problem, called pph , that can be solved in polynomial (even linear) time. Even this may not be fast enough when the whole genome is studied, leading to the question of whether parallel algorithms can be used to solve the pph problem. In the present paper we answer this question affirmatively, but we also give lower complexity bounds on its complexity. In detail, we show that the problem lies in Mod 2 L, a subclass of the circuit complexity class NC2, and is hard for logarithmic space and thus presumably not in NC1. We also investigate variants of the pph problem that have been studied in the literature, like the perfect path phylogeny haplotyping problem and the combined problem where a perfect phylogeny of maximal parsimony is sought, and show that some of these variants are TC0-complete or lie in AC0.

Influence of Tree Topology Restrictions on the Complexity of Haplotyping with Missing Data

Haplotyping, also known as haplotype phase prediction, is the problem of predicting likely haplotypes from genotype data. One fast haplotyping method is based on an evolutionary model where a perfect phylogenetic tree is sought that explains the observed data. Unfortunately, when data entries are missing as is often the case in laboratory data, the resulting incomplete perfect phylogeny haplotyping problem ipph is NP-complete and no theoretical results are known concerning its approximability, fixed-parameter tractability, or exact algorithms for it. Even radically simplified versions, such as the restriction to phylogenetic trees consisting of just two directed paths from a given root, are still NP-complete; but here a fixed-parameter algorithm is known. We show that such drastic and ad hoc simplifications are not necessary to make ipph fixed-parameter tractable: We present the first theoretical analysis of an algorithm, which we develop in the course of the paper, that works for arbitrary instances of ipph . On the negative side we show that restricting the topology of perfect phylogenies does not always reduce the computational complexity: while the incomplete directed perfect phylogeny problem is well-known to be solvable in polynomial time, we show that the same problem restricted to path topologies is NP-complete.

Canonizing Graphs of Bounded Tree Width in Logspace

Graph canonization is the problem of computing a unique representative, a canon, from the isomorphism class of a given graph. This implies that two graphs are isomorphic exactly if their canons are equal. We show that graphs of bounded tree width can be canonized in deterministic logarithmic space (logspace). This implies that the isomorphism problem for graphs of bounded tree width can be decided in logspace. In the light of isomorphism for trees being hard for the complexity class logspace, this makes the ubiquitous classes of graphs of bounded tree width one of the few classes of graphs for which the complexity of the isomorphism problem has been exactly determined.

Phylogeny - and parsimony-based haplotype inference with constraints

Haplotyping, also known as haplotype phase prediction, is the problem of predicting likely haplotypes based on genotype data. One fast computational haplotyping method is based on an evolutionary model where a perfect phylogenetic tree is sought that explains the observed data. In their CPM 2009 paper, Fellows et al. studied an extension of this approach that incorporates prior knowledge in the form of a set of candidate haplotypes from which the right haplotypes must be chosen. While this approach may help to increase the accuracy of haplotyping methods, it was conjectured that the resulting formal problem constrained perfect phylogeny haplotyping might be NP-complete. In the present paper we present a polynomial-time algorithm for it. Our algorithmic ideas also yield new fixed-parameter algorithms for related haplotyping problems based on the maximum parsimony assumption.