Oleksii K. Tyshchenko

An evolving neuro-fuzzy system for online fuzzy clustering

An evolving neuro-fuzzy system and its online learning procedure are proposed in this paper. The system is based on nodes of a special type. A quality estimation process is defined by finding an optimal value of the used cluster validity index.

An evolving connectionist system for data stream fuzzy clustering and its online learning

Abstract An evolving cascade neuro-fuzzy system and its online learning procedure are considered in this paper. The system is based on conventional Kohonen neurons. The proposed system solves a clustering task of non-stationary data streams under uncertainty conditions when data come in the form of a sequential stream in an online mode. A quality estimation process is defined by finding an optimal value of the used cluster validity index.

A cascade deep neuro-fuzzy system for high-dimensional online possibilistic fuzzy clustering

A cascade deep learning system (based on neuro-fuzzy nodes) and its online learning procedure are proposed in this paper. The system is based on nodes of a special type. A goal function of a special type is used for possibilistic high-dimensional fuzzy clustering. To estimate a clustering quality of data processing, an optimal value of a cluster validity index is used.

A deep cascade neuro-fuzzy system for high-dimensional online fuzzy clustering

A deep cascade system (based on neuro-fuzzy nodes) and its online learning procedure are proposed in this paper. A number of layers can grow unlimitedly during a self-learning procedure. The system is based on nodes of a special type. A goal function of a special type is used for probabilistic high-dimensional fuzzy clustering. To assess a clustering quality of data processing, a neurons architecture of a special type is introduced.

A neuro-fuzzy Kohonen network for data stream possibilistic clustering and its online self-learning procedure

Abstract A task of Data Stream Fuzzy Clustering is considered when data is processed sequentially under a priori uncertainty conditions about both a number of clusters and a degree of clusters’ overlapping. A modified two-layer neuro-fuzzy Kohonen network is used for solving the possibilistic fuzzy clustering tasks. This system tunes centers’ coordinates and membership levels of every pattern to clusters during the self-learning procedure and automatically increases a number of neurons during data processing. A distinguishing feature of the proposed approach is its computational simplicity due to the fact that a recurrent modification of the possibilistic fuzzy clustering procedure is used for tuning the network’s parameters. The proposed neuro-fuzzy system is based on the concepts of evolving systems of Computational Intelligence, the recurrent optimization, the possibilistic fuzzy clustering, and Data Stream Mining.

A deep cascade neural network based on extended neo-fuzzy neurons and its adaptive learning algorithm

An architecture and learning methods for deep neural networks that increase a number of layers and adjust their synaptic weights in an online mode are proposed in the article. The systems architecture is based on nodes of a special type (extended neo-fuzzy neurons) which possess enhanced approximating properties. A main feature of the proposed network is a learning process for each node that is performed sequentially in an online mode.

A hybrid growing ENFN-based neuro-fuzzy system and its rapid deep learning

An architecture and learning methods for a growing neuro-fuzzy system that enlarges an amount of layers and tunes their synaptic weights in an online way are introduced in the paper. A structure of the hybrid system is built with the help of extended neo-fuzzy neurons which are characterized by improved approximating capabilities. The main peculiar feature of the introduced system is a learning method for each structural element that is carried out sequentially in an online manner.

A reservoir radial-basis function neural network in prediction tasks

A reservoir radial-basis function neural network, which is based on the ideas of reservoir computing and neural networks and designated for solving extrapolation tasks of nonlinear non-stationary stochastic and chaotic time series under conditions of a short learning sample, is proposed in the paper. The network is built with the help of a radial-basis function neural network with an input layer, which is organized in a special manner and a kernel membership function. The proposed system provides high approximation quality in terms of a mean squared error and a high convergence speed using the second-order learning procedure. A software product that implements the proposed neural network has been developed. A number of experiments have been held in order to research the system’s properties. Experimental results prove the fact that the developed architecture can be used in Data Mining tasks and the fact that the proposed neural network has a higher accuracy compared to traditional forecasting neural systems.

Self-learning Procedures for a Kernel Fuzzy Clustering System

The paper exemplifies several self-learning methods through the prism of diverse objective functions used for training a kernel fuzzy clustering system. A self-learning process for synaptic weights is implemented in terms of the competitive learning concept and the probabilistic fuzzy clustering approach. The main feature of the introduced fuzzy clustering system is its capability to cluster data in an online way under conditions when clusters are rather likely to be of an arbitrary shape (which cannot usually be separated in a linear manner) and to be mutually intersecting. Generally speaking, the offered system’s topology is mainly based on both the fuzzy clustering neural network by Kohonen and the general regression neural network. When it comes to training this hybrid system, it is grounded on both the lazy and optimization-based learning concepts.

A Multidimensional Adaptive Growing Neuro-Fuzzy System and Its Online Learning Procedure

The paper presents learning algorithms for a multidimensional adaptive growing neuro-fuzzy system with optimization of a neuron ensemble in every cascade. A building block for this architecture is a multidimensional neo-fuzzy neuron. The demonstrated system is distinguished from the well-recognized cascade systems in its ability to handle multidimensional data sequences in an online fashion, which makes it possible to treat non-stationary stochastic and chaotic data with the demanded accuracy. The most important privilege of the considered hybrid neuro-fuzzy system is its trait to accomplish a procedure of parallel computation for a data stream based on peculiar elements with upgraded approximating properties. The developed system turns out to be rather easy from the effectuation standpoint; it holds a high processing speed and approximating features. Compared to acclaimed countertypes, the developed system guarantees computational simpleness and owns both filtering and tracking aptitudes. The proposed system, which is ultimately a growing (evolving) system of computational intelligence, assures processing the incoming data in an online fashion just unlike the rest of conventional systems.