Erik Pitzer

A Comprehensive Survey on Fitness Landscape Analysis

In the past, the notion of fitness landscapes has found widespread adoption. Many different methods have been developed that provide a general and abstract framework applicable to any optimization problem. We formally define fitness landscapes, provide an in-depth look at basic properties and give detailed explanations and examples of existing fitness landscape analysis techniques. Moreover, several common test problems or model fitness landscapes that are frequently used to benchmark algorithms or analysis methods are examined and explained and previous results are consolidated and summarized. Finally, we point out current limitations and open problems pertaining to the subject of fitness landscape analysis.

Architecture and Design of the HeuristicLab Optimization Environment

Many optimization problems cannot be solved by classical mathematical optimization techniques due to their complexity and the size of the solution space. In order to achieve solutions of high quality though, heuristic optimization algorithms are frequently used. These algorithms do not claim to find global optimal solutions, but offer a reasonable tradeoff between runtime and solution quality and are therefore especially suitable for practical applications. In the last decades the success of heuristic optimization techniques in many different problem domains encouraged the development of a broad variety of optimization paradigms which often use natural processes as a source of inspiration (as for example evolutionary algorithms, simulated annealing, or ant colony optimization). For the development and application of heuristic optimization algorithms in science and industry, mature, flexible and usable software systems are required. These systems have to support scientists in the development of new algorithms and should also enable users to apply different optimization methods on specific problems easily. The architecture and design of such heuristic optimization software systems impose many challenges on developers due to the diversity of algorithms and problems as well as the heterogeneous requirements of the different user groups. In this chapter the authors describe the architecture and design of their optimization environment HeuristicLab which aims to provide a comprehensive system for algorithm development, testing, analysis and generally the application of heuristic optimization methods on complex problems.

Benefits of plugin-based heuristic optimization software systems

Plugin-based software systems are the next step of evolution in application development. By supporting fine grained modularity not only on the source code but also on the post-compilation level, plugin frameworks help to handle complexity, simplify application configuration and deployment, and enable users or third parties to easily enhance existing applications with self-developed modules without having access to the whole source code. In spite of these benefits, plugin-based software systems are seldom found in the area of heuristic optimization. Some reasons for this drawback are discussed, several benefits of a plugin-based heuristic optimization software system are highlighted and some ideas are shown, how a heuristic optimization meta-model as the basis of a thorough plugin infrastructure for heuristic optimization could be defined.

Towards large-scale sample annotation in gene expression repositories

BackgroundLarge repositories of biomedical research data are most useful to translational researchers if their data can be aggregated for efficient queries and analyses. However, inconsistent or non-existent annotations describing important sample details such as name of tissue or cell line, histopathological type, and subject characteristics like demographics, treatment, and survival are seldom present in data repositories, making it difficult to aggregate data.ResultsWe created a flexible software tool that allows efficient annotation of samples using a controlled vocabulary, and report on its use for the annotation of over 12,500 samples.ConclusionWhile the amount of data is very large and seemingly poorly annotated, a lot of information is still within reach. Consistent tool-based re-annotation enables many new possibilities for large scale interpretation and analyses that would otherwise be impossible.

DSGeo: Software tools for cross-platform analysis of gene expression data in GEO

The Gene Expression Omnibus (GEO) is the largest resource of public gene expression data. While GEO enables data browsing, query and retrieval, additional tools can help realize its potential for aggregating and comparing data across multiple studies and platforms. This paper describes DSGeo-a collection of valuable tools that were developed for annotating, aggregating, integrating, and analyzing data deposited in GEO. The core set of tools include a Relational Database, a Data Loader, a Data Browser, and an Expression Combiner and Analyzer. The application enables querying for specific sample characteristics and identifying studies containing samples that match the query. The Expression Combiner application enables normalization and aggregation of data from these samples and returns these data to the user after filtering, according to the users preferences. The Expression Analyzer allows simple statistical comparisons between groups of data. This seamless integration makes annotated cross-platform data directly available for analysis.

Evaluation of a large-scale biomedical data annotation initiative.

BackgroundThis study describes a large-scale manual re-annotation of data samples in the Gene Expression Omnibus (GEO), using variables and values derived from the National Cancer Institute thesaurus. A framework is described for creating an annotation scheme for various diseases that is flexible, comprehensive, and scalable. The annotation structure is evaluated by measuring coverage and agreement between annotators.ResultsThere were 12,500 samples annotated with approximately 30 variables, in each of six disease categories – breast cancer, colon cancer, inflammatory bowel disease (IBD), rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), and Type 1 diabetes mellitus (DM). The annotators provided excellent variable coverage, with known values for over 98% of three critical variables: disease state, tissue, and sample type. There was 89% strict inter-annotator agreement and 92% agreement when using semantic and partial similarity measures.ConclusionWe show that it is possible to perform manual re-annotation of a large repository in a reliable manner.

Production fine planning using a solution archive of priority rules

Production Fine Planning is often performed directly using all information and assuming that it is fixed. In practice, however, this information changes regularly and the plan has to be adapted. This often means a complete rescheduling of all operations. We present a new approach to this problem by optimizing priority rules that can sort the available next actions. These priority rules often yield similar results even though they do not resemble each other. By using genetic programming to build these priority rules, a distributed system to compute the simulations and a solution archive with a cache of hundreds of thousands of priority rules, new insights into priority rule-based optimization are gained. This archive does not only speed up calculation by avoiding re-simulation of the same rule but can provide a pseudo Pareto front of shorter sub-optimal solutions that facilitate interpretation of the more complex rules and their evolution during the optimization process.

Comprehensive and automatic fitness landscape analysis using heuristiclab

Many different techniques have been developed for fitness landscape analysis. We present a consolidated and uniform implementation inside the heuristic optimization platform HeuristicLab. On top of these analysis methods a new approach to empirical measurement of isotropy is presented that can be used to extend existing methods. Results are shown using the existing implementation within HeuristicLab 3.3.

Integration of flexible interfaces in optimization software frameworks for simulation-based optimization

Optimization of simulation parameters is an important task in many different sciences where simulation is used to model and analyze complex processes and behaviors. In this work it is shown how users, such as researchers, students, and practitioners can benefit from the integration of data-exchange-interfaces in optimization software system. The development of such an interface enables users to couple their own systems and use preimplemented algorithms for their application. The interface description is based on a unified protocol buffer approach which can be ported to further frameworks and optimization software systems. The benefits of a modular architecture, such as in the HeuristicLab optimization environment, will be examined under the light of a successful integration. HeuristicLab is available on the web under the GPL license, its application to the optimization of planning and control systems in manufacturing environments will be shown as a case study in this work. The concrete subject of the case study is a production scenario where different control strategies are used to plan different products. The question is whether machines should be dedicated to a certain control strategy or whether the machines should be shared. The quality is measured by the achieved service level and amounted inventory costs.

Generic hardness estimation using fitness and parameter landscapes applied to robust taboo search and the quadratic assignment problem

Fitness landscape analysis methods have become an increasingly popular topic for research. The future application of these methods to metaheuristics can yield advanced self-adaptive metaheuristics and knowledge bases that can take the role of expert systems in the field of optimization. One important feature of such an expert system would be the prediction of algorithm effort on a certain instance. Estimating whether a certain algorithm is able to tackle the problem adequately or not is a valuable piece of information that currently only an experienced human expert can give. The ability to generate such an advice automatically is, therefore, an important milestone. While fitness landscape analysis methods have been developed for exactly this purpose, it has been shown in the past that single-value analyses have limited applicability. Here, a general method for extracting fitness landscape features will be shown in combination with regression models that indicate a strong correlation between the actual and the predicted effort. Significant potential to increase the prediction quality arises when combining several measures each derived from several different sampling trajectories.