Skip to main content

2018 | Buch

Artificial Intelligence and Soft Computing

17th International Conference, ICAISC 2018, Zakopane, Poland, June 3-7, 2018, Proceedings, Part II

herausgegeben von: Prof. Leszek Rutkowski, Dr. Rafał Scherer, Marcin Korytkowski, Witold Pedrycz, Ryszard Tadeusiewicz, Jacek M. Zurada

Verlag: Springer International Publishing

Buchreihe : Lecture Notes in Computer Science

insite
SUCHEN

Über dieses Buch

The two-volume set LNAI 10841 and LNAI 10842 constitutes the refereed proceedings of the 17th International Conference on Artificial Intelligence and Soft Computing, ICAISC 2018, held in Zakopane, Poland in June 2018.

The 140 revised full papers presented were carefully reviewed and selected from 242 submissions. The papers included in the second volume are organized in the following five parts: computer vision, image and speech analysis; bioinformatics, biometrics, and medical applications; data mining; artificial intelligence in modeling, simulation and control; and various problems of artificial intelligence.

Inhaltsverzeichnis

Frontmatter
Correction to: Analytical Realization of the EM Algorithm for Emission Positron Tomography

In the original version the name of the 4th author was incorrectly stated as “Piotr Filutowicz”. It was corrected to “Zbigniew Filutowicz”.

Robert Cierniak, Piotr Dobosz, Piotr Pluta, Zbigniew Filutowicz

Computer Vision, Image and Speech Analysis

Frontmatter
Moving Object Detection and Tracking Based on Three-Frame Difference and Background Subtraction with Laplace Filter

Moving object detection and tracking is an important research field. Currently, ones of the core algorithms used for tracking include frame difference method (FD), background subtraction method (BS), and optical flow method. Here, authors are looking at the first two approaches since very adequate for very fast real-time treatments whereas optical flow has higher computation cost since related to a dense estimation. Combination of FD and BS with filters and edge detectors is a way to achieve sparse detection fast. This paper presents a tracking algorithm based on a new combination of FD and BS, using Canny edge detector and Laplace filter. Laplace filter occupies a leading role to sharpen the outlines and details. Canny edge detector identifies and extracts edge information. Morphology processing is used to eliminate interfering items finally. Experimental results show that 3FDBD-LC method has higher detection accuracy and better noise suppression than current combination methods on sequence images from standard datasets.

Beibei Cui, Jean-Charles Créput
Robust Lane Extraction Using Two-Dimension Declivity

A new robust lane marking extraction algorithm for monocular vision is proposed based on Two-Dimension Declivity. It is designed for the urban roads with difficult conditions (shadow, high brightness, etc.). In this paper, we propose a locating system which, from an embedded camera, allows lateral positioning of a vehicle by detecting road markings. The primary contribution of the paper is that it supplies a robust method made up of six steps: (i) Image Pre-processing, (ii) Enhanced Declivity Operator (DE), (iii) Mathematical Morphology, (iv) Labeling, (v) Hough Transform and (vi) Line Segment Clustering. The experimental results have shown the high performance of our algorithm in various road scenes. This validation stage has been done with a sequence of simulated images. Results are very promising: more than 90% of marking lines are extracted for less than 12% of false alarm.

Mohamed Fakhfakh, Nizar Fakhfakh, Lotfi Chaari
Segmentation of the Proximal Femur by the Analysis of X-ray Imaging Using Statistical Models of Shape and Appearance

Using image processing to assist in the diagnostic of diseases is a growing challenge. Segmentation is one of the relevant stages in image processing. We present a strategy of complete segmentation of the proximal femur (right and left) in anterior-posterior pelvic radiographs using statistical models of shape and appearance for assistance in the diagnostics of diseases associated with femurs. Quantitative results are provided using the DICE coefficient and the processing time, on a set of clinical data that indicate the validity of our proposal.

Joel Oswaldo Gallegos Guillen, Laura Jovani Estacio Cerquin, Javier Delgado Obando, Eveling Castro-Gutierrez
Architecture of Database Index for Content-Based Image Retrieval Systems

In this paper, we present a novel database index architecture for retrieving images. Effective storing, browsing and searching collections of images is one of the most important challenges of computer science. The design of architecture for storing such data requires a set of tools and frameworks such as relational database management systems. We create a database index as a DLL library and deploy it on the MS SQL Server. The CEDD algorithm is used for image description. The index is composed of new user-defined types and a user-defined function. The presented index is tested on an image dataset and its effectiveness is proved. The proposed solution can be also be ported to other database management systems.

Rafał Grycuk, Patryk Najgebauer, Rafał Scherer, Agnieszka Siwocha
Symmetry of Hue Distribution in the Images

In the paper, a new symmetry measure is proposed to evaluate the symmetry/asymmetry of the hue distribution within the segmented part of the image. A new symmetry/asymmetry area measure (ASM) as well as their parts: the asymmetry measures of: the shape distribution (ASMShape), hue distribution (ASMHue) and structures distribution (ASMStuct) are proposed and discussed. In the paper, a dermatological asymmetry measure in shape (DASMShape) and hue (DASMHue) are presented and discussed thoroughly as well as their ASMShape and ASMHue applications. The hue distribution of the symmetry/asymmetry of the segmented skin lesion is discussed. One of the DASMHue measures is thoroughly presented. The results of the DASMHue algorithm based on the threshold binary masks using PH2 dataset shows stronger overestimating results but the total ratio 95.8% of correctly and overestimated cases is better than the ratio which takes into account only shape alone.

Piotr Milczarski
Image Completion with Smooth Nonnegative Matrix Factorization

Nonnegative matrix factorization is an unsupervised learning method for part-based feature extraction and dimensionality reduction of nonnegative data with a variety of models, algorithms, structures, and applications. Smooth nonnegative matrix factorization assumes the estimated latent factors are locally smooth, and the smoothness is enforced by the underlying model or the algorithm. In this study, we extended one of the algorithms for this kind of factorization to an image completion problem. It is the B-splines ADMM-NMF (Nonnegative Matrix Factorization with Alternating Direction Method of Multipliers) that enforces smooth feature vectors by assuming they are represented by a linear combination of smooth basis functions, i.e. B-splines. The numerical experiments performed on several incomplete images show that the proposed method outperforms the other algorithms in terms of the quality of recovered images.

Tomasz Sadowski, Rafał Zdunek
A Fuzzy SOM for Understanding Incomplete 3D Faces

This paper presents a new recognition method for three-dimensional geometry of the human face. The method measures biometric distances between features in 3D. It relies on the common self-organizing map method with fixed topological distances. It is robust to missing parts of the face due to the introduction of an original fuzzy certainty mask.

Janusz T. Starczewski, Katarzyna Nieszporek, Michał Wróbel, Konrad Grzanek
Feature Selection for ‘Orange Skin’ Type Surface Defect in Furniture Elements

The surfaces of furniture elements having the orange skin surface defect were investigated in the context of selecting optimum features for surface classification. Features selected from a set of 50 features were considered. Seven feature selection methods were used. The results of these selections were aggregated and found consistently positive for some of the features. Among them were primarily the features based on local adaptive thresholding and on Hilbert curves used to evaluate the image brightness variability. These types of features should be investigated further in order to find the features with more significance in the problem of surface quality inspection. The groups of features which appeared least profitable in the analysis were the two features based on percolation, and the one based on Otsu global thresholding.

Bartosz Świderski, Michał Kruk, Grzegorz Wieczorek, Jarosław Kurek, Katarzyna Śmietańska, Leszek J. Chmielewski, Jarosław Górski, Arkadiusz Orłowski
Image Retrieval by Use of Linguistic Description in Databases

In this paper, a new method of image retrieval is proposed. This concerns retrieving color digital images from a database that contains a specific linguistic description considered within the theory of fuzzy granulation and computing with words. The linguistic description is generated by use of the CIE chromaticity color model. The image retrieval is performed in different way depending on users’ knowledge about the color image. Specific database queries can be formulated for the image retrieval.

Krzysztof Wiaderek, Danuta Rutkowska, Elisabeth Rakus-Andersson

Bioinformatics, Biometrics and Medical Applications

Frontmatter
On the Use of Principal Component Analysis and Particle Swarm Optimization in Protein Tertiary Structure Prediction

We discuss applicability of Principal Component Analysis and Particle Swarm Optimization in protein tertiary structure prediction. The proposed algorithm is based on establishing a low-dimensional space where the sampling (and optimization) is carried out via Particle Swarm Optimizer (PSO). The reduced space is found via Principal Component Analysis (PCA) performed for a set of previously found low-energy protein models. A high frequency term is added into this expansion by projecting the best decoy into the PCA basis set and calculating the residual model. Our results show that PSO improves the energy of the best decoy used in the PCA considering an adequate number of PCA terms.

Óscar Álvarez, Juan Luis Fernández-Martínez, Celia Fernández-Brillet, Ana Cernea, Zulima Fernández-Muñiz, Andrzej Kloczkowski
The Shape Language Application to Evaluation of the Vertebra Syndesmophytes Development Progress

In this paper, a measure for assessment the progress of pathological changes in spine bones is introduced. The definition of the measure is based on a syntactic description of geometric features of the bone contours. The proposed approach is applied for analysis of vertebra syndesmophytes in X-ray images of the spine. It turns out that the proposed measure assesses the progress of the disease effectively. The results obtained by the algorithm based on the introduced measure are consistent with the assessment done by an expert.

Marzena Bielecka, Rafał Obuchowicz, Mariusz Korkosz
Analytical Realization of the EM Algorithm for Emission Positron Tomography

The presented paper describes an analytical iterative approach to reconstruction problem for positron emission tomography (PET). The reconstruction problem is formulated taking into consideration the statistical properties of signals obtained by PET scanner and the analytical methodology of image processing. Computer simulations have been performed which prove that the reconstruction algorithm described here, does indeed significantly outperform conventional analytical methods on the quality of the images obtained.

Robert Cierniak, Piotr Dobosz, Piotr Pluta, Zbigniew Filutowicz
An Application of Graphic Tools and Analytic Hierarchy Process to the Description of Biometric Features

AHP is a well-known method supporting decision-making based on a pairwise comparison process. Previous results of our research show that this tool can be effectively used to describe biometric features, in particular facial parts. In this paper, we present an original and innovative development of this approach augmented by a graphical interface that allows the user to get rid of restrictions in the form of certain numerical (linguistic) values, which were adapted beforehand, answering questions about comparisons of individual features. The presented results of experiments show the efficiency and ease of use of AHP based on a graphical interface in a context of description of biometric features. An application a proper non-linear transformation which parameters can be found on a basis of Particle Swarm Optimization can significantly improve the consistency of expert’s evaluation.

Paweł Karczmarek, Adam Kiersztyn, Witold Pedrycz
On Some Aspects of an Aggregation Mechanism in Face Recognition Problems

In the paper, we investigate the problem of an aggregation of classifiers based on numerical and linguistic values of facial features. In the literature, there are many reports of the studies discussing the aggregation or information fusion, however in the situation when the specific classification methods utilize numeric, not linguistic values. Here, we examine the well-known methods (Eigenfaces, Fisherfaces, LBP, MB-LBP, CCBLD) supported by the linguistic values of the measurable facial segments. The detailed results of experiments on the MUCT and PUT facial databases show which of the common aggregation functions and methods have a significant potential to improve the classification process.

Paweł Karczmarek, Adam Kiersztyn, Witold Pedrycz
Nuclei Detection in Cytological Images Using Convolutional Neural Network and Ellipse Fitting Algorithm

Morphometric analysis of nuclei play an essential role in cytological diagnostics. Cytological samples contain hundreds or thousands of nuclei that need to be examined for cancer. The process is tedious and time-consuming but can be automated. Unfortunately, segmentation of cytological samples is very challenging due to the complexity of cellular structures. To deal with this problem, we are proposing an approach, which combines convolutional neural network and ellipse fitting algorithm to segment nuclei in cytological images of breast cancer. Images are preprocessed by the colour deconvolution procedure to extract hematoxylin-stained objects (nuclei). Next, convolutional neural network is performing semantic segmentation of preprocessed image to extract nuclei silhouettes. To find the exact location of nuclei and to separate touching and overlapping nuclei, we approximate them using ellipses of various sizes and orientations. They are fitted using the Bayesian object recognition approach. The accuracy of the proposed approach is evaluated with the help of reference nuclei segmented manually. Tests carried out on breast cancer images have shown that the proposed method can accurately segment elliptic-shaped objects.

Marek Kowal, Michał Żejmo, Józef Korbicz
Towards the Development of Sensor Platform for Processing Physiological Data from Wearable Sensors

The paper outlines a mobile sensor platform aimed at processing physiological data from wearable sensors. We discuss the requirements related to the use of low-cost portable devices in this scenario. Experimental analysis of four such devices, namely Microsoft Band 2, Empatica E4, eHealth Sensor Platform and BITalino (r)evolution is provided. Critical comparison of quality of HR and GSR signals leads to the conclusion that future works should focus on the BITalino, possibly combined with the MS Band 2 in some cases. This work is a foundation for possible applications in affective computing and telemedicine.

Krzysztof Kutt, Wojciech Binek, Piotr Misiak, Grzegorz J. Nalepa, Szymon Bobek
Severity of Cellulite Classification Based on Tissue Thermal Imagining

In this article we present a novel approach to cellulite classification that can be personlised based on non-contact thermal imaging using IR thermography. By analysing the superficial temperature distribution of the body it is possible to diagnose the stages of cellulite development. The study investigates thermal images of posterior of thighs of female volunteers and identifies cellulite areas in an automatic way using image processing. The Growing Bubble Algorithm has been used for thermal picture conversion into valid input vector for a neural network based classifier scheme. Using machine learning process of training the input database was prepared as the stage of cellulite classifier according to the state of the art Nürnberger-Müller diagnosis scheme. Our work demonstrates that it is possible to diagnose the cellulite with over 70% accuracy using a cost-effective, simple and unsophisticated classifier which operates on low-definition pictures. In essence, our work shows that IR-thermography, when coupled with computer aided image analysis and processing, can be a very convenient and effective tool to enable personalized diagnosis and preventive medicine to improve the quality of life of women suffering from cellulite problems.

Jacek Mazurkiewicz, Joanna Bauer, Michal Mosion, Agnieszka Migasiewicz, Halina Podbielska
Features Selection for the Most Accurate SVM Gender Classifier Based on Geometrical Features

In the paper, we have focused on the problem of choosing the best set of features in the task of gender classification/recognition. Choosing a minimum set of features, that can give satisfactory results is also important in the case where only a part of the face is visible. The minimum set of features can simplify the classification process to make it useful for mobile applications. Many authors have used SVM in facial classification and recognition problems, but there are not many works using facial geometry features in the classification neither in SVM. Almost all works are based on the appearance-based methods. In the paper, we show that the classifier constructed on the base of only two or three geometric facial features can give satisfactory (though not always optimal) results with accuracy 82% and positive predictive value 87%, also in incomplete facial images. We show that Matlab and Mathematica can produce very different SVMs given the same data.

Piotr Milczarski, Zofia Stawska, Shane Dowdall
Parallel Cache Efficient Algorithm and Implementation of Needleman-Wunsch Global Sequence Alignment

An approach allowing us to improve the locality of a parallel Needleman-Wunsch (NW) global sequence alignment algorithm is proposed. The original NW algorithm works with an arbitrary gap penalty function and examines all possible gap lengths. To compute the score of an element of an NW array, cells gap symbols are looked back over entire row and column as well as one adjacent cell. We modified the NW algorithm so to read cells only with the row-major order by means of forming a copy of the transposed scoring array. The loop skewing technique is used to generate parallel code. A formal parallel NW algorithm is presented. Experimental results demonstrate super-linear speed-up factor of the accelerated code due to considerable increasing code locality on a studied modern multi-core platform.

Marek Pałkowski, Krzysztof Siedlecki, Włodzimierz Bielecki
Using Fuzzy Numbers for Modeling Series of Medical Measurements in a Diagnosis Support Based on the Dempster-Shafer Theory

This work concern attempts to model imprecise symptoms in the medical diagnosis support tools. Patient’s self-check is very important, particularly in chronic diseases. In hypertension or diabetes patients record measurements. Still, these measurements are made in different circumstances, thus they are imprecise. A physician takes into account rather a trend in a series of measurements to diagnose a patient. Till now, knowledge engineers’ approach is different since they often use a single value as input information of a decision support system. In this work, a series of measurements is modeled as a fuzzy number. The main purpose of the presented approach is to check whether it is possible to replace a single measurement with a series of measurements in the diagnosis support system and to examine the impact of this change on the diagnosis process. Preliminary results show that use of the fuzzy number in determining the diagnosis may increase its certainty and can be profitable when used in real medical problems.

Sebastian Porebski, Ewa Straszecka
Averaged Hidden Markov Models in Kinect-Based Rehabilitation System

In this paper the Averaged Hidden Markov Models (AHMMs) are examined for the upper limb rehabilitation purposes. For the data acquisition the Microsoft Kinect 2.0 sensor is used. The system is intended for low-functioning autistic children whose rehabilitation is often based on sequences of images presenting the subsequent gestures. The number of such training sets is limited and the preparation of a new one is not available for everyone, whereas each child requires the individual therapy. The advantage of the presented system is that new activities models could be easily added.The conducted experiments provide satisfactory results, especially in the case of single hand rehabilitation and both hands rehabilitation based on asymmetric gestures.

Aleksandra Postawka, Przemysław Śliwiński
Genome Compression: An Image-Based Approach

With the advent of Next Generation Sequencing Technologies, it has been possible to reduce the cost and time of genome sequencing. Thus, there was a significant increase in demand for genomes that were assembled daily. This demand requires more efficient techniques for storing and transmitting genomic data. In this research, we discussed the horizontal compression of lossless genomic sequences, using two image formats, WEBP, and FLIF. For this, the genomic sequence is transformed into a matrix of colored pixels, where an RGB color is assigned to each symbol of the A, T, C, G alphabet at a position x-y. The WEBP format showed the best data-rate saving (76.15%, SD = 0.84) when compared to FLIF. In addition, we compared the data-rate savings of two specialized DELIMINATE and MPCompress genomic data compression tools with WEBP. The results obtained show that the WEBP is close to DELIMINATE (76.03%, SD = 2.54%) and MFCompress (76.97%). SD = 1.36%). Finally, we suggest using WEBP for genomic data compression.

Kelvin Vieira Kredens, Juliano Vieira Martins, Osmar Betazzi Dordal, Edson Emilio Scalabrin, Roberto Hiroshi Herai, Bráulio Coelho Ávila
Stability of Features Describing the Dynamic Signature Biometric Attribute

Behavioral biometric attributes tend to change over time. Due to this, analysis of their changes is an important issue in the context of identity verification. In this paper, we present an evaluation of stability of features describing the dynamic signature biometric attribute. The dynamic signature is represented by nonlinear waveforms describing dynamics of the signing process. Our analysis takes into account a set of features extracted using a partitioning of the signature in comparison to so-called global features of the signature. It shows which features change more and how it is associated with identification efficiency. Our simulations were performed using ATVS-SLT DB dynamic signature database.

Marcin Zalasiński, Krzysztof Cpałka, Konrad Grzanek

Data Mining

Frontmatter
Text Categorization Improvement via User Interaction

In this paper, we propose an approach to improvement of text categorization using interaction with the user. The quality of categorization has been defined in terms of a distribution of objects related to the classes and projected on the self-organizing maps. For the experiments, we use the articles and categories from the subset of Simple Wikipedia. We test three different approaches for text representation. As a baseline we use Bag-of-Words with weighting based on Term Frequency-Inverse Document Frequency that has been used for evaluation of neural representations of words and documents: Word2Vec and Paragraph Vector. In the representation, we identify subsets of features that are the most useful for differentiating classes. They have been presented to the user, and his or her selection allow increase the coherence of the articles that belong to the same category and thus are close on the SOM.

Jakub Atroszko, Julian Szymański, David Gil, Higinio Mora
Uncertain Decision Tree Classifier for Mobile Context-Aware Computing

Knowledge discovery from uncertain data is one of the major challenges in building modern artificial intelligence applications. One of the greatest achievements in this area was made with a usage of machine learning algorithms and probabilistic models. However, most of these methods do not work well in systems which require intelligibility, efficiency and which operate on data are not only uncertain but also infinite. This is the most common case in mobile contex-aware computing. In such systems data are delivered in streaming manner, requiring from the learning algorithms to adapt their models iteratively to changing environment. Furthermore, models should be understandable for the user allowing their instant reconfiguration. We argue that all of these requirements can be met with a usage of incremental decision tree learning algorithm with modified split criterion. Therefore, we present a simple and efficient method for building decision trees from infinite training sets with uncertain instances and class labels.

Szymon Bobek, Piotr Misiak
An Efficient Prototype Selection Algorithm Based on Dense Spatial Partitions

In order to deal with big data, techniques for prototype selection have been applied for reducing the computational resources that are necessary to apply data mining approaches. However, most of the proposed approaches for prototype selection have a high time complexity and, due to this, they cannot be applied for dealing with big data. In this paper, we propose an efficient approach for prototype selection. It adopts the notion of spatial partition for efficiently dividing the dataset in sets of similar instances. In a second step, the algorithm extracts a prototype of each of the densest spatial partitions that were previously identified. The approach was evaluated on 15 well-known datasets used in a classification task, and its performance was compared to those of 6 state-of-the-art algorithms, considering two measures: accuracy and reduction. All the obtained results show that, in general, the proposed approach provides a good trade-off between accuracy and reduction, with a significantly lower running time, when compared with other approaches.

Joel Luís Carbonera, Mara Abel
Complexity of Rule Sets Induced by Characteristic Sets and Generalized Maximal Consistent Blocks

We study mining incomplete data sets with two interpretations of missing attribute values, lost values and “do not care” conditions. For data mining we use characteristic sets and generalized maximal consistent blocks. Additionally, we use three types of probabilistic approximations, lower, middle and upper, so altogether we apply six approaches to data mining. Since it was shown that an error rate, associated with such data mining is not universally smaller for any approach, we decided to compare complexity of induced rule sets. Therefore, our objective is to compare six approaches to mining incomplete data sets in terms of complexity of induced rule sets. We conclude that there are statistically significant differences between these approaches.

Patrick G. Clark, Cheng Gao, Jerzy W. Grzymala-Busse, Teresa Mroczek, Rafal Niemiec
On Ensemble Components Selection in Data Streams Scenario with Gradual Concept-Drift

In the paper we study the issue of components selection of an ensemble for data stream classification. Decision about adding or removing single component has significant meaning not only for an accuracy in the current instant, but can be also significant for the further stream processing. The algorithm proposed in this paper is an enhanced version of the ASE (Automatically Sized Ensemble) algorithm which guarantees that a new component will be added to the ensemble only if it increases the accuracy not only for the current data chunk but also for the whole data stream. The algorithm is designed to improve data stream processing in the case when one concept is gradually replaced by the other. The Hellinger distance is applied to allow adding a new component, if its predictions differ significantly from the rest of the ensemble, even though that component does not increase accuracy of the whole ensemble.

Piotr Duda
An Empirical Study of Strategies Boosts Performance of Mutual Information Similarity

In the recent years, the application of mutual information based measures has received broad popularity. The mutual information MINE measure is asserted to be the best strategy for identification of relationships in challenging data sets. A major weakness of the MINE similarity metric concerns its high execution time. To address the performance issue numerous approaches are suggested both with respect to improvement of software implementations and with respect to the application of simplified heuristics. However, none of the approaches manage to address the high execution-time of MINE computation.In this work, we address the latter issue. This paper presents a novel MINE implementation which manages a 530x+ performance increase when compared to established approaches. The novel high-performance approach is the result of a structural evaluation of 30+ different MINE software implementations, implementations which do not make use of simplified heuristics. Hence, the proposed strategy for computation of MINE mutual information is both accurate and fast. The novel mutual information MINE software is available at https://bitbucket.org/oekseth/mine-data-analysis/downloads/. To broaden the applicability the high-performance MINE metric is integrated into the hpLysis machine learning library (https://bitbucket.org/oekseth/hplysis-cluster-analysis-software).

Ole Kristian Ekseth, Svein-Olav Hvasshovd
Distributed Nonnegative Matrix Factorization with HALS Algorithm on Apache Spark

Nonnegative Matrix Factorization (NMF) is a commonly-used unsupervised learning method for extracting parts-based features and dimensionality reduction from nonnegative data. Many computational algorithms exist for updating the latent nonnegative factors in NMF. In this study, we propose an extension of the Hierarchical Alternating Least Squares (HALS) algorithm to a distributed version using the state-of-the-art framework - Apache Spark. Spark gains its popularity among other distributed computational frameworks because of its in-memory approach which works much faster than well-known Apache Hadoop. The scalability and efficiency of the proposed algorithm is confirmed in the numerical experiments, performed on real data as well as synthetic ones.

Krzysztof Fonał, Rafał Zdunek
Dimensionally Distributed Density Estimation

Estimating density is needed in several clustering algorithms and other data analysis methods. Straightforward calculation takes O(N2) because of the calculation of all pairwise distances. This is the main bottleneck for making the algorithms scalable. We propose a faster O(N logN) time algorithm that calculates the density estimates in each dimension separately, and then simply cumulates the individual estimates into the final density values.

Pasi Fränti, Sami Sieranoja
Outliers Detection in Regressions by Nonparametric Parzen Kernel Estimation

A certain observation which is unusual or different from all other ones is called the outlier or anomaly. Appropriate evaluation of data is a crucial problem in modelling of the real objects or phenomena. Actually investigated problems often are based on data mass-produced by computer systems, without careful inspection or screening. The great amount of generated and processed information (e.g. so-called Big-Data) cause that possible outliers often go unnoticed and the result is that they can be masked. However, in regression, this situation can be more complicated. The identification and evaluation of the extremely atypical measurements in observations, for instance in some areas of medicine, geology, particularly in seismology (earthquakes), is precisely the outliers that are the subjects of interest. In this paper, a nonparametric procedure based on Parzen kernel for estimation of unknown function is applied. Evaluation of which measurements in input data-set could be recognized as outliers and possibly should be removed has been performed using the Cook’s Distance formula. Anomaly detection is still an important problem to be researched within diverse areas and application domains.

Tomasz Galkowski, Andrzej Cader
Application of Perspective-Based Observational Tunnels Method to Visualization of Multidimensional Fractals

Methods of multidimensional data visualization are frequently applied in the qualitative analysis allowing to state some properties of this data. They are based only on using the transformation of the multidimensional space into a two-dimensional one which represents the screen in a way ensuring not to lose important properties of the data. Thanks to this it is possible to observe some searched data properties in the most natural way for human beings–through the sense of sight. In this way, the whole analysis is conducted excluding applications of complex algorithms serving to get information about these properties. The example of a multidimensional data visualization method is a relatively new method of perspective-based observational tunnels. It was proved earlier that this method is efficient in the analysis of real data located in a multidimensional space of features obtained by characters recognition. Its efficiency was also shown by the analysis of multidimensional real data describing coal samples. In this paper, another aspect of using this method was shown–to visualize artificially generated five-dimensional fractals located in a five-dimensional space. The purpose of such a visualization can be to obtain views of such multidimensional objects as well as to adapt and teach our mind to percept, recognize and perhaps understand objects of a higher number of dimensions than 3. Our understanding of such multidimensional data could significantly influence the way of perceiving complex multidimensional relations in data and the surrounding world. The examples of obtained views of five-dimensional fractals were shown. Such a fractal looks like a completely different object from different perspectives. Also, views of the same fractal obtained using the PCA, MDS and autoassociative neural networks methods are presented for comparison.

Dariusz Jamroz
Estimation of Probability Density Function, Differential Entropy and Other Relative Quantities for Data Streams with Concept Drift

In this paper estimators of nonstationary probability density function are proposed. Additionally, applying the trapezoidal method of numerical integration, the estimators of two information-theoretic measures are presented: the differential entropy and the Renyi’s quadratic differential entropy. Finally, using an analogous methodology, estimators of the Cauchy-Schwarz divergence and the probability density function divergence are proposed, which are used to measure the differences between two probability density functions. All estimators are proposed in two variants: one with the sliding window and one with the forgetting factor. Performance of all the estimators is verified using numerical simulations.

Maciej Jaworski, Patryk Najgebauer, Piotr Goetzen
System for Building and Analyzing Preference Models Based on Social Networking Data and SAT Solvers

Discovering and modeling preferences has an important meaning in the modern IT systems, also in the intelligent and multi-agent systems which are context sensitive and should be proactive. The preference modelling enables understanding the needs of objects working within intelligent spaces, in an intelligent city. There was presented a proposal for a system, which, based on logical reasoning and using advanced SAT solvers, is able to analyze data from social networks for preference determination in relation to its own presented offers from different domains. The basic algorithms of the system were presented as well as the validation of practical application.

Radosław Klimek
On Asymmetric Problems of Objects’ Comparison

In the paper, we describe selected problems which appear during the process of comparison of the objects. The direction of objects’ comparison seems to have essential role because such comparison may not be symmetric. Thus, we can say that two objects may be viewed as an attempt to determine the degree to which they are similar or different. Asymmetric phenomena of comparing such objects is emphasized and discussed.

Maciej Krawczak, Grażyna Szkatuła
A Recommendation Algorithm Considering User Trust and Interest

A traditional collaborative filtering recommendation algorithm has problems with data sparseness, a cold start and new users. With the rapid development of social network and e-commerce, building the trust between users and user interest tags to provide a personalized recommendation is becoming an important research issue. In this study, we propose a probability matrix factorization model (STUIPMF) by integrating social trust and user interest. First, we identified implicit trust relationship between users and potential interest label from the perspective of user rating. Then, we used a probability matrix factorization model to conduct matrix decomposition of user ratings information, user trust relationship, and user interest label information, and further determined the user characteristics to ease data sparseness. Finally, we used an experiment based on the Epinions website’s dataset to verify our proposed method. The results show that the proposed method can improve the recommendation’s accuracy to some extent, ease a cold start and solve new user problems. Meanwhile, the STUIPMF approach, we propose, also has a good scalability.

Chuanmin Mi, Peng Peng, Rafał Mierzwiak
Automating Feature Extraction and Feature Selection in Big Data Security Analytics

Feature extraction and feature selection are the first tasks in pre-processing of input logs in order to detect cybersecurity threats and attacks by utilizing data mining techniques in the field of Artificial Intelligence. When it comes to the analysis of heterogeneous data derived from different sources, these tasks are found to be time-consuming and difficult to be managed efficiently.In this paper, we present an approach for handling feature extraction and feature selection utilizing machine learning algorithms for security analytics of heterogeneous data derived from different network sensors. The approach is implemented in Apache Spark, using its python API, named pyspark.

Dimitrios Sisiaridis, Olivier Markowitch
Improvement of the Simplified Silhouette Validity Index

The fundamental issue of data clustering is an evaluation of results of clustering algorithms. Lots of methods have been proposed for cluster validation. The most popular approach is based on internal cluster validity indices. Among this kind of indices, the Silhouette index and its computationally simpled version, i.e. the SimplifiedSilhouette, are frequently used. In this paper modification of the SimplifiedSilhouette index is proposed. The suggested approach is based on using an additional component, which improves clusters validity assessment. The performance of the new cluster validity indices has been demonstrated for artificial and real datasets, where the PAM clustering algorithm has been applied as the underlying clustering technique.

Artur Starczewski, Krzysztof Przybyszewski
Feature Extraction in Subject Classification of Text Documents in Polish

In this work we evaluate two different methods for deriving features for a subject classification of text documents. The first method uses the standard Bag-of-Words (BoW) approach, which represents the documents with vectors of frequencies of selected terms appearing in the documents. This method heavily relies on the natural language processing (NLP) tools to properly preprocess text in the grammar- and inflection-conscious way. The second approach is based on the word-embedding technique recently proposed by Mikolov and does not require any NLP preprocessing. In this method the words are represented as vectors in continuous space and this representation of words is used to construct the feature vectors of the documents. We evaluate these fundamentally different approaches in the task of classification of Polish language Wikipedia articles with 34 subject areas. Our study suggests that the word-embedding based features seem to outperform the standard NLP-based features providing sufficiently large training dataset is available.

Tomasz Walkowiak, Szymon Datko, Henryk Maciejewski
Efficiency of Random Decision Forest Technique in Polish Companies’ Bankruptcy Prediction

The purpose of the paper was to compare the accuracy of traditional bankruptcy prediction models with the Random Forest method. In particular, the paper verifies 2 research hypotheses (verification was based on the representative sample of Polish companies): [H1]: The Random Forest algorithm (trained on a representative set of companies) is more accurate than traditional bankruptcy prediction methods: logit and linear discriminant models, and [H2]: The Random Forest algorithm efficiently uses normalized financial statement data (there is no need to calculate financial ratios).

Joanna Wyrobek, Krzysztof Kluza
TUP-RS: Temporal User Profile Based Recommender System

As e-commerce continues to emerge in recent years, online stores compete intensely to improve the quality of recommender systems. However, most existing recommender systems failed to consider both long-term and short-term preferences of users based on purchase behavior patterns, ignoring the fact that requirements of users are dynamic. To this end, we present TUP-RS (Temporal User Profile based Recommender System) in this paper. Specifically, the contributions of this paper are two folds: (i) the long-term and short-term preferences from the topic model are combined to construct the temporal user profiles; (ii) the co-training method which shares the parameters in the same feature space is employed to increase the accuracy. We study a subset of data from Amazon and demonstrate that TUP-RS outperforms state-of-the-art methods. Moreover, our recommendation lists are time-sensitive.

Wanling Zeng, Yang Du, Dingqian Zhang, Zhili Ye, Zhumei Dou
Feature Extraction of Surround Sound Recordings for Acoustic Scene Classification

This paper extends the traditional methodology of acoustic scene classification based on machine listening towards a new class of multichannel audio signals. It identifies a set of new features of five-channel surround recordings for classification of the two basic spatial audio scenes. Moreover, it compares the three artificial intelligence-based classification approaches to audio scene classification. The results indicate that the method based on the early fusion of features is superior compared to those involving the late fusion of signal metrics.

Sławomir K. Zieliński

Artificial Intelligence in Modeling, Simulation and Control

Frontmatter
Cascading Probability Distributions in Agent-Based Models: An Application to Behavioural Energy Wastage

This paper presents a methodology to cascade probabilistic models and agent-based models for fine-grained data simulation, which improves the accuracy of the results and flexibility to study the effect of detailed parameters. The methodology is applied on residential energy consumption behaviour, where an agent-based model takes advantage of probability distributions used in probabilistic models to generate energy consumption of a house with a focus on energy waste. The implemented model is based on large samples of real data and provides flexibility to study the effect of social parameters on the energy consumption of families. The results of the model highlighted the advantage of the cascading methodology and resulted in two domain-specific conclusions: (1) as the number of occupants increases, the family becomes more efficient, and (2) young, unemployed, and part-time occupants cause less energy waste in small families than full-time and older occupants. General insights on how to target families with energy interventions are included at last.

Fatima Abdallah, Shadi Basurra, Mohamed Medhat Gaber
Symbolic Regression with the AMSTA+GP in a Non-linear Modelling of Dynamic Objects

In this paper, we present a new version of the State Transition Algorithm, which allows to automatically determine the number and range of local models that describe the behaviour of a non-linear dynamic object. We used this data as input for genetic programming algorithm in order to create a simple functional model of the non-linear dynamic object which is not computationally demanded and has high accuracy.

Łukasz Bartczuk, Piotr Dziwiński, Andrzej Cader
A Population Based Algorithm and Fuzzy Decision Trees for Nonlinear Modeling

The paper presents a new approach for using the fuzzy decision trees for non-linear modeling based on the capabilities of participle swarm optimization and evolutionary algorithms. The most nonlinear dynamic objects have their approximate nonlinear model. Their parameters are known or can be determined by one of the typical identification procedure. The obtained approximate nonlinear model describes well the identified dynamic object only in the operating point. In this work, we use hybrid model composed with of two parts: approximate nonlinear model and fuzzy decision tree. The fuzzy decision tree contains correction values of the parameters in terminal nodes. The hybrid model ensures sufficient accuracy for the practical applications. A participle swarm optimization and evolutionary algorithm were used for identification of the parameters of the approximate nonlinear model and fuzzy decision tree. An important benefit of the proposed method is the obtained characteristics of the unknown parameters of the approximate nonlinear model described by the terminal nodes of the fuzzy decision tree. They present valuable and interpretable knowledge for the experts concerning the essence of the unknown phenomena.

Piotr Dziwiński, Łukasz Bartczuk, Krzysztof Przybyszewski
The Hybrid Plan Controller Construction for Trajectories in Sobolev Space

This paper proposes a new integrated approach to the hybrid plan controller construction. It forms a synergy of the logic-based approach in terms of LTL-description and automata of Büchi with the integral-based approach. It is shown that the integral-based complementation may be naturally exploited in detection of the robot trajectories by the appropriate control functions.

Krystian Jobczyk, Antoni Ligȩza
Temporal Traveling Salesman Problem – in a Logic- and Graph Theory-Based Depiction

In this paper, a new temporal extension of Traveling Salesman Problem (TSP) – as an old optimization problem – is proposed. This proposal stems from a need to elucidate TSP not only as an optimization problem, but also as a potentially paradigmatic problem for the subject specification of temporal planning. This new Temporal Traveling Salesman Problem is described in two ways – in the graph-based depiction and in terms of logic to be interpreted later by the so-called fibred semantics.

Krystian Jobczyk, Piotr Wiśniewski, Antoni Ligȩza
Modelling the Affective Power of Locutions in a Persuasive Dialogue Game

One of the most important contemporary directions of development in the field of artificial intelligence is to equip AI systems with emotional intelligence. This work is part of this trend. The paper presents a mathematical model that allows us to describe changes in players’ emotional states as a response to dialogue actions. To this end, we use the paradigm of dialogue games and propose a method of rating locutions. The method is inspired by the affective rating system SAM which uses Mehrabian’s PAD space which distinguishes emotions because of three attributes: Pleasantness (valence) (P), Arousal (A), and Dominance (D). Emotions that are analyzed are taken from Ekman’s model with five universally accepted labels: fear, disgust, anger, sadness, and joy. In addition, we describe the emerging tool for the realization of dialogue games with emotional reasoning. This tool is the basis for designing a system for verifying the properties of dialog protocols.

Magdalena Kacprzak, Anna Sawicka, Andrzej Zbrzezny
Determination of a Matrix of the Dependencies Between Features Based on the Expert Knowledge

In the paper, we investigate the problem of replacing long-lasting and expensive research by expert knowledge. The proposed innovative method is a far-reaching improvement of the AHP method. Through the use of a slider, the proposed approach can be used by experts who have not yet met the AHP method or do not feel comfortable when using classic approach related to words and numbers. In the series of experiments, we confirm the efficiency of the method in a modeling of electricity consumption in teleinformatics and in an application of biodiversity to urban planning.

Adam Kiersztyn, Paweł Karczmarek, Khrystyna Zhadkovska, Witold Pedrycz
Dynamic Trust Scoring of Railway Sensor Information

A sensor can encounter many situations where its readings can be untrustworthy and the ability to recognise this is an important and challenging task. It opens the possibility to assess sensors for forensic or maintenance purposes, compare them or fuse their information. We present a proposition to score a piece of information produced by a sensor as an aggregation of three dimensions called reliability, likelihood and credibility into a trust value that take into account a temporal component. The approach is validated on data from the railway domain.

Marcin Lenart, Andrzej Bielecki, Marie-Jeanne Lesot, Teodora Petrisor, Adrien Revault d’Allonnes
Linear Parameter-Varying Two Rotor Aero-Dynamical System Modelling with State-Space Neural Network

In every model-based approaches, i.e., fault diagnosis, fuzzy control, robust fault-tolerant control, the exact model is crucial. This paper presents a methodology which allows to obtain an exact model of high-order, non-linear cross-coupled system, namely Two Rotor Aero-dynamical System (TRAS), using a state-space neural network. Moreover, the resulting model is presented in a linear parameter-varying (LPV) form making it easier to analyze (i.e., its stability and controllability) and control. Such a form is obtained by direct transformation of the neural network structure into quasi-LPV model. For the neural network modelling, a SSNN Toolbox is utilized.

Marcel Luzar, Józef Korbicz
Evolutionary Quick Artificial Bee Colony for Constrained Engineering Design Problems

The Artificial Bee Colony (ABC) is a well-known simple and efficient bee inspired metaheuristic that has been showed to achieve good performance on real valued optimization problems. Inspired by such, a Quick Artificial Bee Colony (QABC) was proposed by Karaboga to enhance the global search and bring better analogy to the dynamic of bees. To improve its local search capabilities, a modified version of it, called Evolutionary Quick Artificial Bee Colony (EQABC), is proposed. The novel algorithm employs the mutation operators found in Evolutionary Strategies (ES) that was applied in ABC from Evolutionary Particle Swarm Optimization (EPSO). In order to test the performance of the new algorithm, it was applied in four large-scale constrained optimization structural engineering problems. The results obtained by EQABC are compared to original ABC, QABC, and ABC + ES, one of the algorithms inspired for the development of EQABC.

Otavio Noura Teixeira, Mario Tasso Ribeiro Serra Neto, Demison Rolins de Souza Alves, Marco Antonio Florenzano Mollinetti, Fabio dos Santos Ferreira, Daniel Leal Souza, Rodrigo Lisboa Pereira

Various Problems of Artificial Intelligence

Frontmatter
Patterns in Video Games Analysis – Application of Eye-Tracker and Electrodermal Activity (EDA) Sensor

The aim of the article is to propose a method for evaluating player’s experience during gameplay using an eye-tracker and galvanic skin response sensor. The method is based on using data obtained from the game, in the light of patterns in game design. The article presents a preliminary, qualitative study, along with an exemplary interpretation of the gameplay of the Hidden Object Puzzle Adventure (HOPA) game.

Iwona Grabska-Gradzińska, Jan K. Argasiński
Improved Behavioral Analysis of Fuzzy Cognitive Map Models

Fuzzy Cognitive Maps (FCMs) are widely applied for describing the major components of complex systems and their interconnections. The popularity of FCMs is mostly based on their simple system representation, easy model creation and usage, and its decision support capabilities.The preferable way of model construction is based on historical, measured data of the investigated system and a suitable learning technique. Such data are not always available, however. In these cases experts have to define the strength and direction of causal connections among the components of the system, and their decisions are unavoidably affected by more or less subjective elements. Unfortunately, even a small change in the estimated strength may lead to significantly different simulation outcome, which could pose significant decision risks. Therefore, the preliminary exploration of model ‘sensitivity’ to subtle weight modifications is very important to decision makers. This way their attention can be attracted to possible problems.This paper deals with the advanced version of a behavioral analysis. Based on the experiences of the authors, their method is further improved to generate more life-like, slightly modified model versions based on the original one suggested by experts. The details of the method is described, its application and the results are presented by an example of a banking application. The combination of Pareto-fronts and Bacterial Evolutionary Algorithm is a novelty of the approach.

Miklós F. Hatwagner, Gyula Vastag, Vesa A. Niskanen, László T. Kóczy
On Fuzzy Sheffer Stroke Operation

The generalization of the classical logical connectives to the fuzzy logic framework has been one of the main research lines since the introduction of fuzzy logic. Although many classical logical connectives have been already generalized, the Sheffer stroke operation has received scant attention. This operator can be used by itself, without any other logical operator, to define a logical formal system in classical logic. Therefore, the goal of this article is to present some initial ideas on the fuzzy Sheffer stroke operation in fuzzy logic. A definition of this operation in the fuzzy logic framework is proposed. Then, a characterization theorem in terms of a fuzzy conjunction and a fuzzy negation is presented. Finally, we show when we can obtain other fuzzy connectives from fuzzy Sheffer stroke operation.

Piotr Helbin, Wanda Niemyska, Pedro Berruezo, Sebastia Massanet, Daniel Ruiz-Aguilera, Michał Baczyński
Building Knowledge Extraction from BIM/IFC Data for Analysis in Graph Databases

This paper deals with the problem of knowledge extraction and processing building related data. Information is retrieved from the IFC files, which are an industry standard for storing building information models (BIM). The IfcWebServer is used as a tool for transforming building information into the graph model. This model is stored in a graph database which allows for obtaining knowledge by defining specific graph queries. The process is illustrated by examples of extracting information needed to find different types of routes in an office building.

Ali Ismail, Barbara Strug, Grażyna Ślusarczyk
A Multi-Agent Problem in a New Depiction

This paper contains a new depiction of the Multi-Agent Problem as motivated by the so-called Nurse Rostering Problem, which forms a workable subcase of this general problem of Artificial Intelligence. Multi-Agent Problem will be presented as a scheduling problem with an additional planning component. The next, the problem will be generalized and different constraints will be put forward. Finally, some workable subcases of Multi-Agent Problem will be implemented in PROLOG-solvers.

Krystian Jobczyk, Antoni Ligȩza
Proposal of a Smart Gun System Supporting Police Interventions

A smart gun idea fits within the smart city concept where the extensive use of information and telecommunication technologies improves working efficiency as well as safety and comfort of residents. There was proposed a system based on Internet of Things (IoT) concept. It detects danger to life and health of policemen and provides them support from other patrols and municipal services. This kind of concept, very different from the previous ones, has already been shown in a prototype version.

Radosław Klimek, Zuzanna Drwiła, Patrycja Dzienisik
Knowledge Representation in Model Driven Approach in Terms of the Zachman Framework

Model driven approach uses distinct models for representing various kinds of complex knowledge. Use of appropriate models allows for taking advantage of formal checking, testing, and validating possibilities available for the models. Although the notations do not provide any design method or software process, this paper offers a step to integrated modeling using them. We present an overview of the existing OMG solutions for knowledge representation used in software engineering, focusing on UML, BPMN, DMN and CMMN models and the diagrams provided by these notations. We perform an analysis of these approaches in terms of Kruchten’s 4+1 view model architecture as well as the Zachman Framework.

Krzysztof Kluza, Piotr Wiśniewski, Antoni Ligęza, Anna Suchenia, Joanna Wyrobek
Rendezvous Consensus Algorithm Applied to the Location of Possible Victims in Disaster Zones

In this paper is presented an alternative to performing the analysis of the sensors in the field of applied cooperative robotics for search and location of disaster victims. This work proposes the use of the Rendezvous algorithm to validate the information coming from the sensors of a multi-robot system. The sensors located in each one of the robotic agents provide a measured value according to the existence or not of victims in the surrounding zone to the robot. Since the information coming from the robots is not the same, however, its belong to the same sensed parameters, the Rendezvous algorithm is used to find a consensus of opinion about the existence of victims. In addition, the swarm of robots uses bio-inspired techniques to generate the navigation algorithm. This navigation algorithm was inspired by the foraging technique used by swarms such as bees, birds or bacteria. The results present some Rendezvous algorithm simulation and robot swarm navigation showing the feasibility of the proposed system.

José León, Gustavo A. Cardona, Luis G. Jaimes, Juan M. Calderón, Pablo Ospina Rodriguez
Exploiting OSC Models by Using Neural Networks with an Innovative Pruning Algorithm

In this paper we have investigated the relationship between the current and the active layer thickness of an organic solar cell (OSC) in order to improve its efficiency by means of a back propagation neural network. In order to preserve the generalization properties of the adopted neural network (NN) in this paper is presented also an innovative pruning technique. The extensive simulations performed show a good agreement between simulated and experimental data with an overall error of about 3%. The obtained results demostrate that the use of an MLP with associated an appropriate pruning algorithm to preserve its generalization capacities permits to accurately reproduce the relationship between the active layer thicknesses and the measured maximum power in an OSC. This neural model can be of great use in manufacturing processes.

Grazia Lo Sciuto, Giacomo Capizzi, Christian Napoli, Rafi Shikler, Dawid Połap, Marcin Woźniak
Critical Analysis of Conversational Agent Technology for Intelligent Customer Support and Proposition of a New Solution

This paper proposes and describes an application of modern, Loebner’s Prize-winning conversational agent technologies in the context of intelligent aid for customer service. The paper defines the requirements, design philosophy and main algorithm for a system that would be able to improve customer service efficiency and functionality using innovative artificial intelligence methods, which can be seen as an increasingly common demand in various branches of business. Emerging problems of interface, knowledge engineering and natural language processing nature are discussed along with proposals of technologies suitable for resolving said issues.

Mateusz Modrzejewski, Przemysław Rokita
Random Forests for Profiling Computer Network Users

In this paper, we present a novel system to detect abnormal behaviour of computer network users based on features of web pages which were requested by a user (e.g. URL address, URL category, the day of week or time when the web page was visited). There are many causes of an abnormal behaviour of network users e.g. a computer can be infected by a virus or a Trojan, a stranger can take control of a computer system, etc. Thus, the proposed system can be a very important security mechanism in networks. The system can be also used to make personal user profiles. We use the bag-of-words model to analyse the text data from firewall logs from 63 users collected over a one and half month period. The 500 GB of the network traffic meta-data allowed to achieve satisfactory classification accuracy.

Jakub Nowak, Marcin Korytkowski, Robert Nowicki, Rafał Scherer, Agnieszka Siwocha
Leader-Follower Formation for UAV Robot Swarm Based on Fuzzy Logic Theory

This paper proposes an algorithm based on a fuzzy logic approach, capable to guide a robot swarm with the aim to keep a leader-follower formation without colliding with other swarm agents. The fuzzy system is programmed and evaluated originally in Matlab, where several experiments were performed. The results depicted a robot swarm showing some bio-inspired behaviors, such as swarm agents surrounding the leader when it is in a static position or when it is traveling from one place to another place. Finally, the proposed fuzzy system was implemented on a drone swarm using V-Rep. The drones simulation shows the swarm navigating together and keeping the leader in the center of the swarm when it is static and following the leader when it is moving. These results could be evaluated in a future work using drone robot swarm in real environments.

Wilson O. Quesada, Jonathan I. Rodriguez, Juan C. Murillo, Gustavo A. Cardona, David Yanguas-Rojas, Luis G. Jaimes, Juan M. Calderón
Towards Interpretability of the Movie Recommender Based on a Neuro-Fuzzy Approach

In the paper, a neuro-fuzzy structure is implemented as a movie recommender. First, a novel method for transforming nominal values of attributes into a numerical form is proposed. This allows representing the nominal values, e.g. movie genres or actors, in a neuro-fuzzy system designed from scratch using the Mendel-Wang algorithm for rules generation. Several experiments illustrate performance of the neuro-fuzzy recommender.

Tomasz Rutkowski, Jakub Romanowski, Piotr Woldan, Paweł Staszewski, Radosław Nielek
Dual-Heuristic Dynamic Programming in the Three-Wheeled Mobile Transport Robot Control

In this work an intelligent discrete tracking control system of a three-wheeled mobile transport robot is presented. The robot is a model of a forklift truck, with a drive wheel mounted in the rear part of the frame in movable steering module. The dynamics of the mobile transport robot was described using the second order Lagrange’s equations. In the tracking control system of the robot the Dual-Heuristic Dynamic Programming algorithm was used, which belongs to the family of Approximate Dynamic Programming algorithms. In the Dual-Heuristic Dynamic Programming algorithm Random Vector Functional Link Neural Networks were used to realize an actor and a critic structure. Numerical tests of robot motion on the desire trajectory were performed. The results of the numerical tests confirmed the correctness of the assumed design assumptions.

Marcin Szuster
Stylometry Analysis of Literary Texts in Polish

In this work we compare different methods for deriving features for text representation in two stylometric tasks of gender and author recognition. The first group of methods uses the Bag-of-Words (BoW) approach, which represents the documents with vectors of frequencies of selected features occurring in the documents. We analyze features such as the most frequent 1000 lemmas, word forms, all lemmas, selected (content insensitive) lemmas, bigrams of grammatical classes and mixture of bigrams of grammatical classes, selected lemmas and punctuations. Moreover, the approach based on the recently proposed fastText algorithm (for vector based representation of text) is also applied. We evaluate these different approaches on two publicly available collections of Polish literary texts from late 19th- and early 20th-century: one consisting of 99 novels from 33 authors and the second one 888 novels from 58 authors. Our study suggests that depending on the corpora the best are the style features (grammatical bigrams) or semantic features (1000 lemmas extracted from the training set). We also noticed the importance of proper division of corpora into training and testing sets.

Tomasz Walkowiak, Maciej Piasecki
Constraint-Based Identification of Complex Gateway Structures in Business Process Models

In this paper, we present a method for identifying parallel and alternative gateway structures in BPMN models. It can be applied in the composition of business processes from their declarative specifications. Our approach is based on a directed graph representation of a business process as well as the constraint programming technique. Provided the information about process activities and relations between them, the proposed approach consists in finding a structure of logical data-based gateways that satisfies the set of predefined constraints. A detailed illustration of our method is preceded by a brief description of BPMN and its formal representation.

Piotr Wiśniewski, Antoni Ligęza
Developing a Fuzzy Knowledge Base and Filling It with Knowledge Extracted from Various Documents

The article describes the process of developing a fuzzy knowledge base. The content of the fuzzy knowledge base is the result of extracting knowledge from the set of documents by subject area. Set of documents consists of the wiki-resources, UML-diagrams, documents and source code of projects. Knowledge base based on the graph database Neo4j. An attempt to implement the mechanism of inference by the contents of a graph database was made. This mechanism is used to generate the screen forms of the user interface dynamically. The contexts allow representing the content of the fuzzy knowledge base in space and time. Each space context is assigned a linguistic label, for example, low, middle, high. This label determines the competence of the expert in the given subject area. Time contexts allow storing the history of the knowledge base content changes. It allows returning to a specific state of the contents of the knowledge base.

Nadezhda Yarushkina, Vadim Moshkin, Aleksey Filippov, Gleb Guskov
Backmatter
Metadaten
Titel
Artificial Intelligence and Soft Computing
herausgegeben von
Prof. Leszek Rutkowski
Dr. Rafał Scherer
Marcin Korytkowski
Witold Pedrycz
Ryszard Tadeusiewicz
Jacek M. Zurada
Copyright-Jahr
2018
Electronic ISBN
978-3-319-91262-2
Print ISBN
978-3-319-91261-5
DOI
https://doi.org/10.1007/978-3-319-91262-2