Grzegorz Baron

Influence of Data Discretization on Efficiency of Bayesian Classifier for Authorship Attribution

Abstract Authorship attribution is one of the research areas in data mining domain and various methods can be employed for performing that task. The paper presents results of research on influence of data discretization on efficiency of Naive Bayes classifier. The analysis has been carried on datasets founded on texts of two male and two female authors using the WEKA data mining software framework. The binary classification was performed separately for both datasets for wide range of parameters of discretization process in order to investigate dependency between ways of discretization and quality of classification using Naive Bayes method. The numerical results of tests have been compared and discussed and some observations and conclusions formulated.

On Approaches to Discretization of Datasets Used for Evaluation of Decision Systems

The paper describes research on ways of datasets discretization, when test datasets are used for evaluation of a classifier. Three different approaches of processing for training and test datasets are presented: “independent”—where discretization is performed separately for both sets assuming that the same algorithm parameters are used; “glued”—where both sets are concatenated, discretized, and resulting set is separated to obtain training and test sets, and finally “test on learn”—where test dataset is discretized using ranges obtained from learning data. All methods have been investigated and tested in authorship attribution domain using Naive Bayes classifier.

Comparison of Cross-Validation and Test Sets Approaches to Evaluation of Classifiers in Authorship Attribution Domain

The presented paper addresses problem of evaluation of decision systems in authorship attribution domain. Two typical approaches are cross-validation and evaluation based on specially created test datasets. Sometimes preparation of test sets can be troublesome. Another problem appears when discretization of input sets is taken into account. It is not obvious how to discretize test datasets. Therefore model evaluation method not requiring test sets would be useful. Cross-validation is the well-known and broadly accepted method, so the question arose if it can deliver reliable information about quality of prepared decision system. The set of classifiers was selected and different discretization algorithms were applied to obtain method invariant outcomes. The comparative results of experiments performed using cross-validation and test sets approaches to system evaluation, and conclusions are presented.

Gesture and voice driven mobile tribot robot using Kinect sensor

The Project is a three devices Microsoft Kinect, Mobile Lego NXT Robot integrated system - and computer. These are controlled in a single application which enables the user to control the NXT Robot remotely by tracking calibration using voice commands.

Remote control of the artificial arm model using 3D hand tracking

Modern sensing technologies create new possibilities to control mobile robots without any dedicated manipulators. In this article authors present a novel method that enables driving of the Mindstorms NXT artificial arm with Microsoft Kinect, using gesture recognition. To imitate movement of an artificial robotic arm, an algorithm of the human-computer interaction is employed using skeleton tracking and gesture control in 3D space.

Analysis of Multiple Classifiers Performance for Discretized Data in Authorship Attribution

In authorship attribution domain single classifiers are often employed in research as elements of decision system. On the other hand, there is intuitive prediction that the use of multiple classifier with fusion of their outcomes may improve the quality of the investigated system. Additionally, discretization can be applied for input data which can be beneficial for the classification accuracy. The paper presents performance analysis of some multiple classifiers basing on the majority voting rule. Ensembles were composed from eight single well known classifiers. Influence of different discretization methods on the quality of the analyzed systems was also investigated.

On Influence of Representations of Discretized Data on Performance of a Decision System

When discretization is used for preprocessing datasets in a decision system different representations of data can be taken into consideration. Typical approach is to use data as it is returned by discretizer, namely as nominal values. But in specific cases such form of data cannot be utilized by next modules of the decision system. Then the possible solution is to convert nominal data again into a numerical form. The paper presents comparison of such approaches applied for different classifiers in stylometry domain.

Application of System on Chip (SoC) devices for the design of a smoke detector working with megapixel image streams

The paper discusses the application of System on Chip devices for processing Megapixel video streams. The domain of image processing using high resolution images is very demanding in the scope of calculating power and frequently exploits special processing hardware. The progress of integration technology brings about SoC which are capable of meeting such processing demands. Characteristics of FPGA and DSP based systems are assessed. A concise survey of chip solutions is presented and a pair of representative systems is chosen for comparison of their properties. The design of a smoke detection system is used for evaluating the advantages and drawbacks of using such SoC solutions. Almost identical image processing algorithms are implemented in the case of FPGA and DSP. Both solutions meet the image processing requirements and are suitable for real time processing. FPGA based solution is much faster as it uses pipelining for improving the utilisation of scarce memory resources.

Touch-Less Java Based Hand Tracking System to Control an Artificial Arm

Modern sensing technologies create new possibilities to control mobile robots without any dedicated manipulators. In this article authors present a novel method that enables driving of the Mindstorms NXT artificial arm with Microsoft Kinect, using gesture recognition and hand tracking. To imitate the movement of an artificial robotic arm, an algorithm of the human-computer interaction is employed using skeleton tracking and gesture control in 3D space. The solution is implemented using free, open source software, Java based.

Method of contour and its characteristic feature extraction from a noisy image

A new method of contour and its characteristic feature extraction form noisy image is presented. For image smoothing, the image convolution with a Gaussian function has been used. Then an inflexion point of an image function in the direction of the image gradient vector has been used as a new contour definition. Behavior of this image gradient gives also the information about spatial orientation of contour segments used as invariant characteristic features. Thus, the angles between segments have been calculated by means of operations on the histogram of contour angles. The method application on a highly noised image has been used for illustration of proper method works.