Jakub Nalepa

Support vector machines training data selection using a genetic algorithm

This paper presents a new method for selecting valuable training data for support vector machines (SVM) from large, noisy sets using a genetic algorithm (GA). SVM training data selection is a known, however not extensively investigated problem. The existing methods rely mainly on analyzing the geometric properties of the data or adapt a randomized selection, and to the best of our knowledge, GA-based approaches have not been applied for this purpose yet. Our work was inspired by the problems encountered when using SVM for skin segmentation. Due to a very large set size, the existing methods are too time-consuming, and random selection is not effective because of the set noisiness. In the work reported here we demonstrate how a GA can be used to optimize the training set, and we present extensive experimental results which confirm that the new method is highly effective for real-world data.

Wrist Localization in Color Images for Hand Gesture Recognition

In this paper we present an extensive study of a two-stage algorithm for wrist localization in color images, which is an important and challenging, yet not extensively studied, step in gesture recognition systems. In the first stage of the algorithm, a color hand image is subject to skin segmentation. Secondly, the wrist is localized in a corresponding binarized skin probability map. In our two-stage approach, the algorithms for both localization stages can be developed and compared separately. Here, we compare our propagation-based skin segmentation algorithm and real-time wrist localization algorithm with other state-of-the-art approaches based on our set of 414 color hand images using two independent sets of ground-truth data.

Adaptive memetic algorithm for minimizing distance in the vehicle routing problem with time windows

This paper presents an adaptive memetic algorithm to solve the vehicle routing problem with time windows (VRPTW). It is a well-known NP-hard discrete optimization problem with two objectives—to minimize the number of vehicles serving a set of geographically dispersed customers, and to minimize the total distance traveled in the routing plan. Although memetic algorithms have been proven to be extremely efficient in solving the VRPTW, their main drawback is an unclear tuning of their numerous parameters. Here, we introduce the adaptive memetic algorithm (AMA-VRPTW) for minimizing the total travel distance. In AMA-VRPTW, a population of solutions evolves with time. The parameters of the algorithm, including the selection scheme, population size and the number of child solutions generated for each pair of parents, are adjusted dynamically during the search. We propose a new adaptive selection scheme to balance the exploration and exploitation of the solution space. Extensive experimental study performed on the well-known Solomon’s and Gehring and Homberger’s benchmark sets confirms the efficacy and convergence capabilities of the proposed AMA-VRPTW. We show that it is very competitive compared with other state-of-the-art techniques. Finally, the influence of the proposed adaptive schemes on the AMA-VRPTW behavior and performance is investigated in a thorough sensitivity analysis. This analysis is complemented with the two-tailed Wilcoxon test for verifying the statistical significance of the results.

A memetic algorithm to select training data for support vector machines

In this paper we propose a new memetic algorithm (MASVM) for fast and efficient selection of a valuable training set for support vector machines (SVMs). This is a crucial step especially in case of large and noisy data sets, since the SVM training has high time and memory complexity. The majority of state-of-the-art methods exploit the data geometry analysis, both in the input and kernel space. Although evolutionary algorithms have been proven to be very efficient for this purpose, they have not been extensively studied so far. Here, we propose a new method employing an adaptive genetic algorithm enhanced by some refinement techniques. The refinements are based on utilizing a pool of the support vectors identified so far at various steps of the algorithm. Extensive experimental study performed on the well-known benchmark, real-world and artificial data sets clearly confirms the efficacy, robustness and convergence capabilities of the proposed approach, and shows that it is competitive compared with other state-of-the-art techniques.

Adaptive memetic algorithm enhanced with data geometry analysis to select training data for SVMs

Support vector machines (SVMs) are one of the most popular and powerful machine learning techniques, but suffer from a significant drawback of the high time and memory complexities of their training. This issue needs to be endured especially in the case of large and noisy datasets. In this paper, we propose a new adaptive memetic algorithm (PCA2MA) for selecting valuable SVM training data from the entire set. It helps improve the classifier score, and speeds up the classification process by decreasing the number of support vectors. In PCA2MA, a population of reduced training sets undergoes the evolution, which is complemented by the refinement procedures. We propose to exploit both a priori information about the training set-extracted using the data geometry analysis-and the knowledge attained dynamically during the PCA2MA execution to enhance the refined sets. Also, we introduce a new adaptation scheme to control the pivotal algorithm parameters on the fly, based on the current search state. Extensive experimental study performed on benchmark, real-world, and artificial datasets clearly confirms the efficacy and convergence capabilities of the proposed approach. We demonstrate that PCA2MA is highly competitive compared with other state-of-the-art techniques. HighlightsWe propose a new adaptive memetic algorithm to select SVM training data.We perform the PCA-based preprocessing to determine valuable training samples.We apply a new scheme to adapt the refined training set size and selection scheme.We evaluate the importance of particular components of our algorithm.We demonstrate the effectiveness and efficiency of the proposed memetic algorithm.

Skin Detection and Segmentation in Color Images

This chapter presents an overview of existing methods for human skin detection and segmentation. First of all, the skin color modeling schemes are outlined, and their limitations are discussed based on the presented experimental study. Then, we explain the techniques which were reported helpful in improving the efficacy of color-based classification, namely (1) textural features extraction, (2) model adaptation schemes, and (3) spatial analysis of the skin blobs. The chapter presents meaningful qualitative and quantitative results obtained during our study, which demonstrate the benefits of exploiting particular techniques for improving the skin detection outcome.

Self-adaptive algorithm for segmenting skin regions

In this paper, we introduce a new self-adaptive algorithm for segmenting human skin regions in color images. Skin detection and segmentation is an active research topic, and many solutions have been proposed so far, especially concerning skin tone modeling in various color spaces. Such models are used for pixel-based classification, but its accuracy is limited due to high variance and low specificity of human skin color. In many works, skin model adaptation and spatial analysis were reported to improve the final segmentation outcome; however, little attention has been paid so far to the possibilities of combining these two improvement directions. Our contribution lies in learning a local skin color model on the fly, which is subsequently applied to the image to determine the seeds for the spatial analysis. Furthermore, we also take advantage of textural features for computing local propagation costs that are used in the distance transform. The results of an extensive experimental study confirmed that the new method is highly competitive, especially for extracting the hand regions in color images.

Skin detection using spatial analysis with adaptive seed

This paper introduces a new method for adaptive skin detection in color images combined with spatial analysis of skin pixels. It has been reported in many works that adaptation of a skin color model to a particular image may decrease the false positives, however the false negatives are considerably high unless a local model is combined with the global one. Another possibility for improvement is to analyze spatial properties of the pixels classified as skin, but this operation strongly depends on the seed extraction technique. Our contribution lies in using a local dynamic skin model learned from the detected faces to extract seeds for the spatial analysis. We present an extensive experimental study confirming that our method outperforms alternative skin detection techniques.

Particle swarm optimization for hyper-parameter selection in deep neural networks

Deep neural networks (DNNs) have achieved unprecedented success in a wide array of tasks. However, the performance of these systems depends directly on their hyper-parameters which often must be selected by an expert. Optimizing the hyper-parameters remains a substantial obstacle in designing DNNs in practice. In this work, we propose to select them using particle swarm optimization (PSO). Such biologically-inspired approaches have not been extensively exploited for this task. We demonstrate that PSO efficiently explores the solution space, allowing DNNs of a minimal topology to obtain competitive classification performance over the MNIST dataset. We showed that very small DNNs optimized by PSO retrieve promising classification accuracy for CIFAR-10. Also, PSO improves the performance of existing architectures. Extensive experimental study, backed-up with the statistical tests, revealed that PSO is an effective technique for automating hyper-parameter selection and efficiently exploits computational resources.

New Selection Schemes in a Memetic Algorithm for the Vehicle Routing Problem with Time Windows

This paper presents an extensive study on the pre- and post-selection schemes in a memetic algorithm (MA) for solving the vehicle routing problem with time windows. In the MA, which is a hybridization of the genetic and local optimization algorithms, the population of feasible solutions evolves with time. The fitness of the individuals is measured based on the fleet size and the total distance traveled by the vehicles servicing a set of geographically scattered customers. Choosing the proper selection schemes is crucial to avoid the premature convergence of the search, and to keep the balance between the exploration and exploitation during the search. We propose new selection schemes to handle these issues. We present how the various selection schemes affect the population diversity, convergence of the search and solutions quality. The quality of the solutions is measured as their proximity to the best currently-known feasible solutions. We present the experimental results for the well-known Gehring and Homberger’s benchmark tests.