The Shortest Superstring Problem, SSP, asks for a shortest string which contains each string from a given set of strings. The problem arises in DNA-sequencing and data compression. We describe a family of three-stage algorithms for SSP and a learning algorithm LA that learns the algorithm from the family which is the most accurate for a given input domain for SSP. We experimentally test the learning algorithm and the evolved algorithms on the input domains originating from various DNA sequences and random domains. The experiments show that learning yields ecient algorithms for SSP that are more than 0.999-accurate. The data suggest that a successfully sequenced DNA strand can be used to signicantly speed up the sequencing of other DNA strands. 1 Introduction The instance of the Shortest Superstring Problem, SSP (sometimes termed SCS, for Shortest Common Superstring), is a nite collection S = fs i g of strings written in a nite alphabet. The problem is to construct a short...

Abstract. The Tutte polynomial of a graph, also known as the partition function of the q-state Potts model, is a 2-variable polynomial graph invariant of considerable importance in both combinatorics and statistical physics. It contains several other polynomial invariants, such as the chromatic polynomial and flow polynomial as partial evaluations, and various numerical invariants such as the number of spanning trees as complete evaluations. However despite its ubiquity, there are no widely-available effective computational tools able to compute the Tutte polynomial of a general graph of reasonable size. In this paper we describe the implementation of a program that exploits isomorphisms in the computation tree to extend the range of graphs for which it is feasible to compute their Tutte polynomials. We also consider edge-selection heuristics which give good performance in practice. We empirically demonstrate the utility of our program on random graphs. More evidence of its usefulness arises from our success in finding counterexamples to a conjecture of Welsh on the location of the real flow roots of a graph. 1

Experimental Algorithmics is concerned with the design, implementation, tuning, debugging and performance analysis of computer programs for solving algorithmic problems. It provides methodologies and tools for designing, developing and experimentally analyzing efficient algorithmic codes and aims at integrating and reinforcing traditional theoretical approaches for the design and analysis of algorithms and data structures. In this paper we survey some relevant contributions to the field of Experimental Algorithmics and we discuss significant examples where the experimental approach helped in developing new ideas, in assessing heuristics and techniques, and in gaining a deeper insight about existing algorithms. 1

Absolute fast converging phylogenetic reconstruction methods are provably guaranteed to recover the true tree with high probability from sequences that grow only polynomially in the number of leaves, once the edge lengths are bounded arbitrarily from above and below. Only a few methods have been determined to be absolute fast converging; these have all been developed in just the last few years, and most are polynomial time. In this paper, we compare pre-existing fast converging methods as well as some new polynomial time methods that we have developed. Our study, based upon simulating evolution under a wide range of model conditions, establishes that our new methods outperform both neighbor joining and the previous fast converging methods, returning very accurate large trees, when these other methods do poorly.

Algorithm Engineering is concerned with the design, analysis, implementation, tuning, debugging and experimental evaluation of computer programs for solving algorithmic problems. It provides methodologies and tools for developing and engineering e#cient algorithmic codes and aims at integrating and reinforcing traditional theoretical approaches for the design and analysis of algorithms and data structures.

Abstract. We address the problem of estimating the degree to which the evolutionary history of a set of molecular sequences violates a strong molecular clock hypothesis. We quantify this deviation formally, by defining the “stretch ” of a model tree with respect to the underlying ultrametric tree (indicated by time). We then define the “minimum stretch ” of a dataset for a tree and show how this can be computed optimally in polynomial time. We also present a polynomial-time algorithm for computing a lower bound on the stretch of a given dataset for any tree. We then explore the performance of standard techniques in systematics for estimating the deviation of a dataset from a molecular clock. We show that standard methods, whether based upon maximum parsimony or maximum likelihood, can return infeasible values (i.e. values for the stretch which cannot be realized on a tree), and often under-estimate the true stretch. This suggests that current approximations of the degree to which data sets deviate from a molecular clock may significantly underestimate these deviations. We conclude with some suggestions for further research.

For automated surveillance applications, estimating the symbol rate of an unknown digital communication signal is an important step in the analysis process. Several papers have investigated using the wavelet transform in symbol rate estimation algorithms. Due to its complexity, closed form analysis of performance is often limited, and simulations may not include practical factors such as carrier frequency offset or symbol pulse shaping. This paper uses an automated statistically based test framework to investigate the performance of the wavelet transform against PSK signals with parameters that span a realistic portion of the High Frequency (HF) signal space. The analysis identifies signal and algorithm parameters that affect performance. We also demonstrate that accurate metrics for estimating the probability of failure/success under realistic operating conditions are available for the db6 wavelet.

In manufacturing control, machine scheduling research has mostly dealt with problems either without maintenance or with deterministic maintenance when no failure can occur. This can be unrealistic in practical settings. In this work, an experimental model is developed to evaluate the effect of corrective and preventive maintenance schemes on scheduling performance in the presence machine failure where the scheduling objective is to minimize schedule duration. We show that neither scheme is clearly superior, but that the applicability of each depends on several system parameters as well as the scheduling environment itself. Further, we show that parameter values can be chosen for which preventive maintenance does better than corrective maintenance. The results provided in this study can be useful to practitioners and to system or machine administrators in manufacturing and elsewhere.

Conducting computational experiments and analyzing their results in a sound manner can be tedious. Experiments have to be organized, i. e. algorithms in various configurations, with several inputs and repetitions have to be run and results have to be analyzed from different perspectives, including statistical evaluation. In this paper we discuss properties of an ideal testbed for experiments and the statistical evaluation thereof, which automates recurring tasks and supports scientific soundness. We present a prototypical testbed, which will realize some of these ideas and which is short of being completed.

Experimental algorithmics yields large amounts of data that depends on many parameters. This paper collects a number of rules for presenting this data in concise, meaningful, understandable diagrams that have sufficiently high quality to be printed in scientific journals. The focus