Todor Petkov

Modeling Logic Gates and Circuits with Generalized Nets

In this paper, modeling of logic gates is presented for the first time. Four models of Generalized Nets (GN)—AND gate, a binary to decimal decoder, delay type flip-flop, n-bit binary counter and logical circuits are presented in the following paper. Here we also suggest using the recently proposed approach of InterCriteria Analysis, based on index matrices and intuitionistic fuzzy sets, which aim to detect possible correlations between pairs of criteria. We can perform the measurements, if we have a set of several logical circuits that can be used to obtain identical output data. The aforementioned logical circuits must be composed of different logical elements. By using several measurement points and different schematics, we can suggest the best solution for the considered type of task.

Generalized net model of encrypting message in an image using self organizing map neural network

This paper describes the combination of artificial neural networks, image processing and encryption for the purpose of encrypting text in a message using Self Organizing Map neural network. The main goal is to send a message between two users, which is encrypted in an image, and if a wrong person receives it he will not be permitted to read the message. The neural network is trained by a training set and divided into twenty-six clusters, where each cluster responds to a letter from the English alphabet. When the procedure is done a random image is applied to the network for testing in order to find the area of alphabetical letters. The process of encryption is described with Generalized net.

Image to Sound Encryption Using a Self-organizing Map Neural Network

This paper describes the process of encrypting image in a sound using artificial neural network. In order to achieve it the process is divided into several steps where each of the steps is described with a generalized net. The main goal is to send an image which is encrypted into a sound between two persons and if a wrong person receives it he will not be permitted to see the image. The neural network is divided into 5 clusters where each cluster responds to areas where the image has to be encrypted. When the procedure ends a random sound is applied to the network for testing and depending on which cluster it enters the necessarily areas are taken and the image is applied on them.

A Hybrid Approach for Modular Neural Network Design Using Intercriteria Analysis and Intuitionistic Fuzzy Logic

Intercriteria analysis (ICA) is a new method, which is based on the concepts of index matrices and intuitionistic fuzzy sets, aiming at detection of possible correlations between pairs of criteria, expressed as coefficients of the positive and negative consonance between each pair of criteria. Here, the proposed method is applied to study the behavior of one type of neural networks, the modular neural networks (MNN), that combine several simple neural models for simplifying a solution to a complex problem. They are a tool that can be used for object recognition and identification. Usually the inputs of the MNN can be fed with independent data. However, there are certain limits when we may use MNN, and the number of the neurons is one of the major parameters during the implementation of the MNN. On the other hand, a high number of neurons can slow down the learning process, which is not desired. In this paper, we propose a method for removing part of the inputs and, hence, the neurons, which in addition leads to a decrease of the error between the desired goal value and the real value obtained on the output of the MNN. In the research work reported here the authors have applied the ICA method to the data from real datasets with measurements of crude oil probes, glass, and iris plant. The method can also be used to assess the independence of data with good results.

A Generalized Net Model of the Neocognitron Neural Network

In this paper a generalized net model of the Neocognitron neural network is presented. A Network Neocognitron is a self-organizing network with the ability to recognize patterns based on the difference of their form. A neocognitron is able to correctly identify an image, even if there is a violation or movement into position. Self-organization in the neocognitron is also realized uncontrollably - training for self-organizing neocognitron takes only a collection of recurring patterns in the recognizable image and does not need the information for categories that include templates. The output producing process is presented by a Generalized net model.

Generalized Net Modelling of the Intuitionistic Fuzzy Evaluation of the Quality Assurance in Universities

In the paper is proposed a method for evaluation of the quality assurance in universities and scientific organizations. The evaluation of the quality is based on criteria, which measure different aspects of university activities and consists of different sub-criteria. For the assessment the theory of intuitionistic fuzzy sets is used. The obtained intuitionistic fuzzy estimations reflect the degree of each criterion’ satisfaction, and non-satisfaction. We also consider a degree of uncertainty that represents such cases wherein is no information about sub-criteria of the current criterion. The generalized model gives possibility for algorithmization of the methodology of forming the quality evaluations is constructed. It provides the possibility for the algorithmization of the process of forming the evaluation of the quality assurance in universities.

Encrypting message in a sound using self organizing map neural network described by a generalized net

This paper describes the process of encrypting message in a sound using artificial neural network. In order to accomplish the task, the process is divided into several steps where each of the steps is described with Generalized net. The main goal is to send an encrypted in a sound message between two persons and if a wrong person receives it he will not be permitted to read the message. The neural network is divided into 27 clusters where each cluster responds to a letter from the English alphabet, the 27th cluster responds to the interval between separate words. When the procedure ends, a random sound is applied to the network for the purpose of finding those areas that correspond to the desirable message.

Generalized Net Model of Optimization of the Self-Organizing Map Learning Algorithm

This paper describes an optimization of the algorithm of self-organizing map neural network. The proposed algorithm takes place during the learning trial. We take into consideration the number of the epochs so their number needs to be decreased. In order to do that, for each epoch the distance from each cluster unit to all training vectors is measured. If the total distance is the same as the distance estimated from the previous epoch, it is assumed that the network is trained and the learning trial stops. The process of optimization is described with the generalized net.

Generalized Net Model of Person Recognition Using ART2 Neural Network and Viola-Jones Algorithm

In this paper we present a method for the purpose to detect a certain person in an image. We use the tools of neural networks and face recognition algorithm to achieve our goal. The type of neural network is unsupervised adaptive resonance theory 2 (ART2). It is trained by the set of person images and divided into two clusters—the first cluster represents the human who has to be found and the second one represents the other people. The algorithm which is used for face detection is Viola-Jones and the combination with neural networks helps to identify the person. The generalized net model is used to describe the recognition process.

A Generalized Net Model Based on Fast Learning Algorithm of Unsupervised Art2 Neural Network

In this paper the fast learning algorithm of unsupervised adaptive resonance theory ART2 neural network is described. At the beginning of the process the algorithm is illustrated step by step by mathematical formulas and it is shown how individual vector changes its values during the training. The network supports clustering by using competitive learning, normalization and suppression of the noise. At the end of the process we have stable recognition clusters with values according to the vectors.