J. Kukal’s research while affiliated with University of Chemistry and Technology, Prague and other places

Human scent traces are often the only pieces of evidence providing information about individuals that were present at a crime scene. In this study, the possibility of sex differentiation using detailed chemical analyses of human scent samples for forensic purposes is discussed. The human scent samples were analyzed through the use of headspace-gas chromatography/mass spectrometry (HS-GC/MS). The results of these chemical analyses were evaluated using several data processing approaches (Linear Support Vector Machine, Quadratic Discriminant Analysis, and Ridge Regression), which were applied to distinguish between sexes from the human scent samples obtained from the palms of six volunteers for twelve weeks. This study indicates that sex differentiation based on the chemical analysis of human scent samples using HS-GC/MS is possible. The best results were obtained using the Ridge Regression with thresholding providing accuracy and a critical sensitivity of the sex differentiation of better than 91%.

Our paper presents a novel approach to pattern classification. The general disadvantage of a traditional classifier is in too different behaviour and optimal parameter settings during training on a given pattern set and the following cross-validation. We describe the term critical sensitivity, which means the lowest reached sensitivity for an individual class. This approach ensures a uniform classification quality for individual class classification. Therefore, it prevents outlier classes with terrible results. We focus on the evaluation of critical sensitivity, as a quality criterion. Our proposed classifier eliminates this disadvantage in many cases. Our aim is to present that easily formed hidden classes can significantly contribute to improving the quality of a classifier. Therefore, we decided to propose classifier will have a relatively simple structure. The proposed classifier structure consists of three layers. The first is linear, used for dimensionality reduction. The second layer serves for clustering and forms hidden classes. The third one is the output layer for optimal cluster unioning. For verification of the proposed system results, we use standard datasets. Cross-validation performed on standard datasets showed that our critical sensitivity-based classifier provides comparable sensitivity to reference classifiers.

The traditional self-organized map (SOM) is learned by Kohonen learning and the most common 2-dimensional grids defining the structure of the map are the hexagonal grid and the rectangular grid. A novel model of self-organization is based on hexagonal grid and diffusion modeling in continuous space which is a good approximation of endorphins propagation and nitric oxide generation in the real brain. Therefore the structure of the system is described by neuron coordinates instead of neighborhood relationships in traditional SOM. The discussed neuron activation using the diffusion process and novel diffusive learning algorithm is based on this activation mentioned above. The novel structure and algorithm are demonstrated on simple examples and real economic applications.

The cluster analysis is a traditional tool for multi-varietal data processing. Using the k-means method, we can split a pattern set into a given number of clusters. These clusters can be used for the final classification of known output classes. This paper focuses on various approaches that can be used for an optimal union of hidden classes. The resulting tasks include binary programming or convex optimization ones. Another possibility of obtaining hidden classes is designing imperfect classifier system. Novel context out learning approach is also discussed as possibility of using simple classifiers as background of the system of hidden classes which are easy to union to output classes using the optimal algorithm. All these approaches are useful in many applications, including econometric research. There are two main methodologies: supervised and unsupervised learning based on given pattern set with known or unknown output classification. Preferring supervised learning, we can combine the context out learning with optimal union of hidden classes to obtain the final classifier. But if we prefer unsupervised learning, we will begin with cluster analysis or another similar approach to also obtain the hidden class system for future optimal unioning. Therefore, the optimal union algorithm is widely applicable for any kind of classification tasks. The presented techniques are demonstrated on an artificial pattern set and on real data related to crisis prediction based on the clustering of macroeconomic indicators.

Novel context out learning approach is discussed as possibility of using simple classifiers which is background of hidden class system. There are two ways how to perform final classification. Having a lot of hidden classes we can build their unions using binary optimization task. Resulting system has the best possible sensitivity over all output classes. Another way is to perform second level linear classification as referential approach. The presented techniques are demonstrated on traditional iris flower task.

Alzheimer’s Disease (AD) is the most frequent form of degenerative dementia and its early diagnosis is essential for effective treatment. Functional imaging modalities including Single Photon Emission Computed Tomography (SPECT) are often used with such an aim. However, conventional evaluation of SPECT images relies on manual reorientation and visual evaluation of tomographic slices which is time consuming, subjective and therefore prone to error. Our aim is to show an automatic Computer-Aided Diagnosis (CAD) system for improving the early detection of the AD. For this purpose, affine invariant descriptors of 3D SPECT image can be useful. The method consists of four steps: evaluation of invariant descriptors obtained using spherical harmonic analysis, statistical testing of their significance, application of regularized binary index models, and model verification via leave-one-out cross-validation scheme. The second approach is based on Support Vector Machine (SVM) classifier and visualization with use of self-organizing maps. Our approaches were tested on SPECT data from 11 adult patients with definite Alzheimer’s disease and 10 adult patients with Amyotrophic Lateral Sclerosis (ALS) who were used as controls. A significant difference between SPECT spherical cuts of AD group and ALS group was both visually and numerically evaluated. c CTU FTS 2018

There are many techniques of image enhancement. Their parameters are traditionally tuned by maximization of SNR criterion, which is unfortunately based on the knowledge of an ideal image. Our approach is based on Hartley entropy, its estimation, and differentiation. Resulting gradient of entropy is estimated without knowledge of ideal images, and it is a subject of minimization. Both SNR maximization and gradient magnitude minimization cause various settings of the given filter. The optimum settings are compared, and their differences are discussed.

The paper deals with principal component analysis and data whitening. The research is done in the area of main economic indicators. This means the data preprocessing problem. The main aim of this paper is to present and discuss the possible ways of data preprocessing. The paper deals with four main approaches. There are compared the results from raw data, absolute differences, relative differences and logarithmic differences. The classic principal component analysis is also used with some improvement, there is described the basement of data whitening. The main aim is to get the good data visualisation. The next aim of such approach can be to identify the similarities between some states and their main trends. For this reason there is presented the comparison of states of Visegrad Group. At this moment there is no aim to deeply discuss the reasons of development in detail. This paper suggests new point of view to time series connected to economic development. The deep analysis of all relationships is the topic for further research.