Johannes Mayer

Lattice-based adaptive random testing

Adaptive Random Testing (ART) denotes a family of testing algorithms that have a better performance compared to pure random testing with respect to the number of test cases necessary to detect the first failure. Many of these algorithms, however, are not very efficient regarding runtime. A new ART algorithm is presented that has a better performance than all other ART methods for the block failure pattern. Its runtime is linear in the number of test cases selected, which is nearly as efficient as pure random testing, as opposed to most other ART methods. This new ART algorithm selects the test cases based on a lattice.

A unified simulation framework for spatial stochastic models

Abstract For spatial stochastic models, a lot of programs exist which deal with the simulation of specific models. But, combining them is not that easy and usually requires greater effort. This paper presents an object-oriented framework, i.e. a set of collaborating abstract and concrete classes, dealing with the simulation of such models. The selected fundamental models are only illustrating examples for the general concept. From the Java implementation of this framework, two code examples are shown, which could also be implemented similarly in any other object-oriented programming language. All interfaces and a lot of concrete classes can be implemented dimension-independent. The design and implementation problems arising in the context of static and dynamic plus sampling are specifically discussed.

An empirical analysis and comparison of random testing techniques

Testing with randomly generated test inputs, namely Random Testing, is a strategy that has been applied succefully in a lot of cases. Recently, some new adaptive approaches to the random generation of test cases have been proposed. Whereas there are many comparisons of Random Testing with Partition Testing, a systematic comparison of random testing techniques is still missing. This paper presents an empirical analysis and comparison of all random testing techniques from the field of Adaptive Random Testing (ART). The ART algorithms are compared for effectiveness using the mean F-measure, obtained through simulation and mutation analysis, and the P-measure. An interesting connection between the testing effectiveness measures F-measure and P-measure is described. The spatial distribution of test cases is determined to explain the behavior of the methods and identify possible shortcomings. Besides this, both the theoretical asymptotic runtime and the empirical runtime for each method are given.

An Empirical Study on the Selection of Good Metamorphic Relations

Full software test automation requires automated test input generation, execution, and output evaluation. The latter task is non-trivial and usually referred to as the oracle problem in software testing. The present paper describes an empirical study on metamorphic testing, an approach to the oracle problem. This study was conducted with common Java implementations of determinant computation in order to evaluate the usefulness of the metamorphic testing approach and to establish general criteria that can be used to quickly assess metamorphic relations with respect to their suitability. The latter is very important, since metamorphic testing is based on so-called metamorphic relations on input-output tuples, which can easily be found. It is, however, crucial to evaluate these relations according to their usefulness. The empirical study enables us to derive general rules that can be used to quickly assess metamorphic relations and identify those that should be considered and studied in more detail with other methods (e.g. with mutation analysis)

Statistical Metamorphic Testing Testing Programs with Random Output by Means of Statistical Hypothesis Tests and Metamorphic Testing

Testing software with random output is a challenging task as the output corresponding to a given input differs from execution to execution. Therefore, the usual approaches to software testing are not applicable to randomized software. Instead, statistical hypothesis tests have been proposed for testing those applications. To apply these statistical hypothesis tests, either knowledge about the theoretical values of statistical characteristics of the program output (e. g. the mean) or a reference implementation (e. g. a legacy system) are required to apply statistical hypothesis tests. But often, both are not available. In the present paper, it is discussed how a testing method called Metamorphic Testing can be used to construct statistical hypothesis tests without knowing exact theoretical characteristics or having a reference implementation. For that purpose, two or more independent output sequences are generated by the implementation under test (IUT). Then, these sequences are compared according to the metamorphic relation using statistical hypothesis tests.

Quality of Software Architectures and Software Quality

Keynotes.- Reexamining the Role of Interactions in Software Architecture.- Are Successful Test Cases Useless or Not?.- QoSA Long Papers.- DoSAM - Domain-Specific Software Architecture Comparison Model.- An Architecture-Centric Approach for Producing Quality Systems.- A Model-Oriented Framework for Runtime Monitoring of Nonfunctional Properties.- Predicting Mean Service Execution Times of Software Components Based on Markov Models.- An XML-Based Language to Support Performance and Reliability Modeling and Analysis in Software Architectures.- Formal Definition of Metrics Upon the CORBA Component Model.- The Architects Dilemma - Will Reference Architectures Help?.- Architectural Reuse in Software Systems In-house Integration and Merge - Experiences from Industry.- Supporting Security Sensitive Architecture Design.- Exploring Quality Attributes Using Architectural Prototyping.- On the Estimation of Software Reliability of Component-Based Dependable Distributed Systems.- Empirical Evaluation of Model-Based Performance Prediction Methods in Software Development.- SOQUA Long Papers.- Automatic Test Generation for N-Way Combinatorial Testing.- Automated Generation and Evaluation of Dataflow-Based Test Data for Object-Oriented Software.- Automated Model-Based Testing of ? Simulation Models with TorX.- Jartege: A Tool for Random Generation of Unit Tests for Java Classes.- FlexTest: An Aspect-Oriented Framework for Unit Testing.- Quality Assurance in Performance: Evaluating Mono Benchmark Results.

Classification of Prostatic Carcinoma with Artificial Neural Networks Using Comparative Genomic Hybridization and Quantitative Stereological Data

Staging of prostate cancer is a mainstay of treatment decisions and prognostication. In the present study, 50 pT2N0 and 28 pT3N0 prostatic adenocarcinomas were characterized by Gleason grading, comparative genomic hybridization (CGH), and histological texture analysis based on principles of stereology and stochastic geometry. The cases were classified by learning vector quantization and support vector machines. The quality of classification was tested by cross-validation. Correct prediction of stage from primary tumor data was possible with an accuracy of 74-80% from different data sets. The accuracy of prediction was similar when the Gleason score was used as input variable, when stereological data were used, or when a combination of CGH data and stereological data was used. The results of classification by learning vector quantization were slightly better than those by support vector machines. A method is briefly sketched by which training of neural networks can be adapted to unequal sample sizes per class. Progression from pT2 to pT3 prostate cancer is correlated with complex changes of the epithelial cells in terms of volume fraction, of surface area, and of second-order stereological properties. Genetically, this progression is accompanied by a significant global increase in losses and gains of DNA, and specifically by increased numerical aberrations on chromosome arms 1q, 7p, and 8p.

Distributional properties of the typical cell of stationary iterated tessellations

Abstract.Distributional properties are considered of the typical cell of stationary iterated tessellations (SIT), which are generated by stationary Poisson-Voronoi tessellations (SPVT) and stationary Poisson line tessellations (SPLT), respectively. Using Neveu’s exchange formula, the typical cell of SIT can be represented by those cells of its component tessellation hitting the typical cell of its initial tessellation. This provides a simulation algorithm without consideration of limits in space. It has been applied in order to estimate the probability densities of geometric characteristics of the typical cell of SIT generated by SPVT and SPLT. In particular, the probability densities of the number of vertices, the perimeter, and the area of the typical cell of such SIT have been determined.

Adaptive random testing by bisection with restriction

Random Testing is a strategy to select test cases based on pure randomness. Adaptive Random Testing (ART), a family of algorithms, improves pure Random Testing by taking common failure pattern into account. The best—in terms of the number of test cases necessary to detect the first failure—ART algorithms, however, are too runtime inefficient. Therefore, a modification of a fast, but not so good ART algorithm, namely ART by Bisection, is presented. This modification requires much less test cases than the original method while retaining its computational efficiency.

On Random Testing of Image Processing Applications

Testing image processing applications is a non-trivial task. Complex inputs have to be generated and complex test results have to be evaluated. In the present paper, models for random generation of images are proposed and compared. The study for their comparison uses mutants of one particular implementation of an image processing operator, namely an implementation of the Euclidean distance transform. Metamorphic relations, necessary properties, and special values are furthermore identified for this distance transform to enable automatic evaluation of test results. These criteria are also compared using mutation analysis. Based on the results, general hints are given on how to choose random models and automatically evaluate test results for testing in the field of image processing