Jeroen Eggermont

Genetic Programming for data classification: partitioning the search space

When Genetic Programming is used to evolve decision trees for data classification, search spaces tend to become extremely large. We present several methods using techniques from the field of machine learning to refine and thereby reduce the search space sizes for decision tree evolvers. We will show that these refinement methods improve the classification performance of our algorithms.

Adaptive Genetic Programming Applied to New and Existing Simple Regression Problems

In this paper we continue our study on adaptive genetic programming. We use Stepwise Adaptation of Weights (SAW) to boost performance of a genetic programming algorithm on simple symbolic regression problems. We measure the performance of a standard GP and two variants of SAW extensions on two different symbolic regression problems from literature. Also, we propose a model for randomly generating polynomials which we then use to further test all three GP variants.

Mixed integer evolution strategies for parameter optimization

Evolution strategies (ESs) are powerful probabilistic search and optimization algorithms gleaned from biological evolution theory. They have been successfully applied to a wide range of real world applications. The modern ESs are mainly designed for solving continuous parameter optimization problems. Their ability to adapt the parameters of the multivariate normal distribution used for mutation during the optimization run makes them well suited for this domain. In this article we describe and study mixed integer evolution strategies (MIES), which are natural extensions of ES for mixed integer optimization problems. MIES can deal with parameter vectors consisting not only of continuous variables but also with nominal discrete and integer variables. Following the design principles of the canonical evolution strategies, they use specialized mutation operators tailored for the aforementioned mixed parameter classes. For each type of variable, the choice of mutation operators is governed by a natural metric for this variable type, maximal entropy, and symmetry considerations. All distributions used for mutation can be controlled in their shape by means of scaling parameters, allowing self-adaptation to be implemented. After introducing and motivating the conceptual design of the MIES, we study the optimality of the self-adaptation of step sizes and mutation rates on a generalized (weighted) sphere model. Moreover, we prove global convergence of the MIES on a very general class of problems. The remainder of the article is devoted to performance studies on artificial landscapes (barrier functions and mixed integer NK landscapes), and a case study in the optimization of medical image analysis systems. In addition, we show that with proper constraint handling techniques, MIES can also be applied to classical mixed integer nonlinear programming problems.

Evolving Fuzzy Decision Trees with Genetic Programming and Clustering

In this paper we present a new fuzzy decision tree representation for data classification using genetic programming. The new fuzzy representation utilizes fuzzy clusters for handling continuous attributes. To make optimal use of the fuzzy classifications of this representation an extra fitness measure is used. The new fuzzy representation will be compared, using several machine learning data sets, to a similar non-fuzzy representation as well as to some other evolutionary and non-evolutionary algorithms from literature.

Raising the Dead: Extending Evolutionary Algorithms with a Case-Based Memory

In dynamically changing environments, the performance of a standard evolutionary algorithm deteriorates. This is due to the fact that the population, which is considered to contain the history of the evolutionary process, does not contain enough information to allow the algorithm to react adequately to changes in the fitness landscape. Therefore, we added a simple, global case-based memory to the process to keep track of interesting historical events. Through the introduction of this memory and a storing and replacement scheme we were able to improve the reaction capabilities of an evolutionary algorithm with a periodically changing fitness function.

Automatic detection of bioresorbable vascular scaffold struts in intravascular optical coherence tomography pullback runs

Bioresorbable vascular scaffolds (BVS) have gained significant interest in both the technical and clinical communities as a possible alternative to metallic stents. For accurate BVS analysis, intravascular optical coherence tomography (IVOCT) is currently the most suitable imaging technique due to its high resolution and the translucency of polymeric BVS struts for near infrared light. However, given the large number of struts in an IVOCT pullback run, quantitative analysis is only feasible when struts are detected automatically. In this paper, we present an automated method to detect and measure BVS struts based on their black cores in IVOCT images. Validated using 3 baseline and 3 follow-up data sets, the method detected 93.7% of 4691 BVS struts correctly with 1.8% false positives. In total, the Dices coefficient for BVS strut areas was 0.84. It concludes that this method can detect BVS struts accurately and robustly for tissue coverage measurement, malapposition detection, strut distribution analysis or 3D scaffold reconstruction.

Mixed-Integer NK landscapes

NK landscapes (NKL) are stochastically generated pseudo-boolean functions with N bits (genes) and K interactions between genes. By means of the parameter K ruggedness as well as the epistasis can be controlled. NKL are particularly useful to understand the dynamics of evolutionary search. We extend NKL from the traditional binary case to a mixed variable case with continuous, nominal discrete, and integer variables. The resulting test function generator is a suitable test model for mixed-integer evolutionary algorithms (MI-EA) – i. e. instantiations of evolution algorithms that can deal with the aforementioned variable types. We provide a comprehensive introduction to mixed-integer NKL and characteristics of the model (global/local optima, computation, etc.). Finally, a first study of the performance of mixed-integer evolution strategies on this problem family is provided, the results of which underpin its applicability for optimization algorithm design.

Mixed-integer optimization of coronary vessel image analysis using evolution strategies

In this paper we compare Mixed-Integer Evolution Strategies (MI-ES)and standard Evolution Strategies (ES)when applied to find optimal solutions for artificial test problems and medical image processing problems. MI-ES are special instantiations of standard ES that can solve optimization problems with different objective variable types (continuous, integer, and nominal discrete). Artificial test problems are generated with a mixed-integer test generator.The practical image processing problem iss the detection of the lumen boundary in IntraVascular UltraSound (IVUS)images. Based on the experimental results, it is shown that MI-ES generally perform better than standard ES on both artifical and practical image processing problems. Moreover it is shown that MI-ES can effectively improve the parameters settings for the IVUS lumen detection algorithm.

Bioresorption and Vessel Wall Integration of a Fully Bioresorbable Polymeric Everolimus-Eluting Scaffold: Optical Coherence Tomography, Intravascular Ultrasound, and Histological Study in a Porcine Model With 4-Year Follow-Up.

OBJECTIVES The aim of the present study was to investigate the relationship between the integration process and luminal enlargement with the support of light intensity (LI) analysis on optical coherence tomography (OCT), echogenicity analysis on intravascular ultrasound, and histology up to 4 years in a porcine model. BACKGROUND In pre-clinical and clinical studies, late luminal enlargement has been demonstrated at long-term follow-up after everolimus-eluting poly-l-lactic acid coronary scaffold implantation. However, the time relationship and the mechanistic association with the integration process are still unclear. METHODS Seventy-three nonatherosclerotic swine that received 112 Absorb scaffolds were evaluated in vivo by OCT, intravascular ultrasound, and post-mortem histomorphometry at 3, 6, 12, 18, 24, 30, 36, 42, and 48 months. RESULTS The normalized LI, which is the signal densitometry on OCT of a polymeric strut core normalized by the vicinal neointima, was able to differentiate the degree of connective tissue infiltration inside the strut cores. Luminal enlargement was a biphasic process at 6 to 18 months and at 30 to 42 months. The latter phase occurred with vessel wall thinning and coincided with the advance integration process demonstrated by the steep change in normalized LI (0.26 [interquartile range (IQR): 0.20 to 0.32] at 30 months versus 0.68 [IQR: 0.58 to 0.83] at 42 months, p < 0.001). CONCLUSIONS In this pre-clinical model, late luminal enlargement relates to strut integration into the arterial wall. Quantitative LI analysis on OCT could be used as a surrogate method for monitoring the integration process of poly-l-lactic acid scaffolds, which could provide insight and understanding on the imaging-related characteristics of the bioresorption process of polylactide scaffolds in human.

Mixed-Integer evolution strategies and their application to intravascular ultrasound image analysis

This paper discusses Mixed-Integer Evolution Strategies and their application to an automatic image analysis system for IntraVascular UltraSound (IVUS) images. Mixed-Integer Evolution Strategies can optimize different types of decision variables, including continuous, nominal discrete, and ordinal discrete values. The algorithm is first applied to a set of test problems with scalable ruggedness and dimensionality. The algorithm is then applied to the optimization of an IVUS image analysis system. The performance of this system depends on a large number of parameters that – so far – need to be chosen manually by a human expert. It will be shown that a mixed-integer evolution strategy algorithm can significantly improve these parameters compared to the manual settings by the human expert.