Cüneyt Güzeliş

A new design method for the complex-valued multistate Hopfield associative memory

A method to store each element of an integral memory set M subset {1,2,...,K}/sup n/ as a fixed point into a complex-valued multistate Hopfield network is introduced. The method employs a set of inequalities to render each memory pattern as a strict local minimum of a quadratic energy landscape. Based on the solution of this system, it gives a recurrent network of n multistate neurons with complex and symmetric synaptic weights, which operates on the finite state space {1,2,...,K}/sup n/ to minimize this quadratic functional. Maximum number of integral vectors that can be embedded into the energy landscape of the network by this method is investigated by computer experiments. This paper also enlightens the performance of the proposed method in reconstructing noisy gray-scale images.

Analysis of input-output clustering for determining centers of RBFN

The key point in design of radial basis function networks is to specify the number and the locations of the centers. Several heuristic hybrid learning methods, which apply a clustering algorithm for locating the centers and subsequently a linear leastsquares method for the linear weights, have been previously suggested. These hybrid methods can be put into two groups, which will be called as input clustering (IC) and input-output clustering (IOC), depending on whether the output vector is also involved in the clustering process. The idea of concatenating the output vector to the input vector in the clustering process has independently been proposed by several papers in the literature although none of them presented a theoretical analysis on such procedures, but rather demonstrated their effectiveness in several applications. The main contribution of this paper is to present an approach for investigating the relationship between clustering process on input-output training samples and the mean squared output error in the context of a radial basis function netowork (RBFN). We may summarize our investigations in that matter as follows: 1) A weighted mean squared input-output quantization error, which is to be minimized by IOC, yields an upper bound to the mean squared output error. 2) This upper bound and consequently the output error can be made arbitrarily small (zero in the limit case) by decreasing the quantization error which can be accomplished through increasing the number of hidden units.

Automatic detection of epileptiform events in EEG by a three-stage procedure based on artificial neural networks

This paper introduces a three-stage procedure based on artificial neural networks for the automatic detection of epileptiform events (EVs) in a multichannel electroencephalogram (EEG) signal. In the first stage, two discrete perceptrons fed by six features are used to classify EEG peaks into three subgroups: 1) definite epileptiform transients (ETs); 2) definite non-ETs; and 3) possible ETs and possible non-ETs. The pre-classification done in the first stage not only reduces the computation time but also increases the overall detection performance of the procedure. In the second stage, the peaks falling into the third group are aimed to be separated from each other by a nonlinear artificial neural network that would function as a postclassifier whose input is a vector of 41 consecutive sample values obtained from each peak. Different networks, i.e., a backpropagation multilayer perceptron and two radial basis function networks trained by a hybrid method and a support vector method, respectively, are constructed as the postclassifier and then compared in terms of their classification performances. In the third stage, multichannel information is integrated into the system for contributing to the process of identifying an EV by the electroencephalographers (EEGers). After the integration of multichannel information, the overall performance of the system is determined with respect to EVs. Visual evaluation, by two EEGers, of 19 channel EEG records of 10 epileptic patients showed that the best performance is obtained with a radial basis support vector machine providing an average sensitivity of 89.1%, an average selectivity of 85.9%, and a false detection rate (per hour) of 7.5.

Stability analysis of generalized cellular neural networks

A rather general class of neural networks, called generalized cellular neural networks (CNNs), is introduced. the new model covers most of the known neural network architectures, including cellular neural networks, Hopfield networks and multilayer perceptrons. Several sets of conditions ensuring the input-output stability and global asymptotic stability of generalized CNNs have been obtained. the conditions for the stability of individual cells are checked in the frequency domain, while the stability of the overall network is analysed in terms of the stability of individual cells and the connectivity characteristics. the results on the global asymptotic stability are useful for the design of a generalized CNN such that the orbit of each state converges to a globally asymptotically stable equilibrium point which depends only on the input and not on the initial state. Such a network defines an algebraic map from the space of external inputs to the space of steady state values of the outputs and hence can accomplish cognitive and computational tasks.

Automatic recognition of sleep spindles in EEG by using artificial neural networks

Abstract In this paper, we introduce a two-stage procedure based on artificial neural networks for the automatic recognition of sleep spindles (SSs) in a multi-channel electroencephalographic signal. In the first stage, a discrete perceptron is used to eliminate definite non-SSs. The pre-classification done in the first stage not only reduces the computation time but also increases the overall detection performance of the procedure. In the second stage, the remaining SS candidates after pre-classification procedure are aimed to be separated from each other by an artificial neural network that would function as a post-classifier. Two different networks, i.e. a backpropagation multilayer perceptron and radial basis support vector machine (SVM), are proposed as the post-classifier and compared in terms of their classification performances. Visual evaluation, by two electroencephalographers (EEGers), of 19 channel EEG records of 6 subjects showed that the best performance is obtained with a radial basis SVM providing an average sensitivity of 94.6% and an average false detection rate of 4.0%.

A canonical representation for piecewise-affine maps and its applications to circuit analysis

A canonical representation for a rather general class of piecewise-affine maps has been developed. This canonical representation extends the canonical representation proposed by L.O. Chua and S.M. Hang (Proc. IEEE, vol.65, no.6, p.915-29, 1977) into PWA (piecewise-affine) partitions, which arise frequently in driving-point, transfer characteristics, and state equations. Thus, a universal canonical representation that is capable of characterizing circuit equations, state equations, and driving-point and transfer characteristics of piecewise-affine circuits in a compact global analytic form has been obtained. The canonical forms developed seem to be promising tools for computational purposes as well as for analytical studies of PWA circuits. >

A new facial expression recognition based on curvelet transform and online sequential extreme learning machine initialized with spherical clustering

In this paper, a novel algorithm is proposed for facial expression recognition by integrating curvelet transform and online sequential extreme learning machine (OSELM) with radial basis function (RBF) hidden node having optimal network architecture. In the proposed algorithm, the curvelet transform is firstly applied to each region of the face image divided into local regions instead of whole face image to reduce the curvelet coefficients too huge to classify. Feature set is then generated by calculating the entropy, the standard deviation and the mean of curvelet coefficients of each region. Finally, spherical clustering (SC) method is employed to the feature set to automatically determine the optimal hidden node number and RBF hidden node parameters of OSELM by aim of increasing classification accuracy and reducing the required time to select the hidden node number. So, the learning machine is called as OSELM-SC. It is constructed two groups of experiments: The aim of the first one is to evaluate the classification performance of OSELM-SC on the benchmark datasets, i.e., image segment, satellite image and DNA. The second one is to test the performance of the proposed facial expression recognition algorithm on the Japanese Female Facial Expression database and the Cohn-Kanade database. The obtained experimental results are compared against the state-of-the-art methods. The results demonstrate that the proposed algorithm can produce effective facial expression features and exhibit good recognition accuracy and robustness.

Patient oriented and robust automatic liver segmentation for pre-evaluation of liver transplantation

Identifying liver region from abdominal computed tomography-angiography (CTA) data sets is one of the essential steps in evaluation of transplantation donors prior to the hepatic surgery. However, due to gray level similarity of adjacent organs, injection of contrast media and partial volume effects; robust segmentation of the liver is a very difficult task. Moreover, high variations in liver margins, different image characteristics with different CT scanners and atypical liver shapes make the segmentation process even harder. In this paper, we propose a three stage (i.e. pre-processing, classification, post-processing); automatic liver segmentation algorithm that adapts its parameters according to each patient by learning the data set characteristics in parallel to segmentation process to address all the challenging aspects mentioned above. The efficiency in terms of the time requirement and the overall segmentation performance is achieved by introducing a novel modular classification system consisting of a K-Means based simple classification system and an MLP based complex one which are combined with a data-dependent and automated switching mechanism that decides to apply one of them. Proposed approach also makes the design of the overall classification system fully unsupervised that depends on the given CTA series only without requiring any given training set of CTA series. The segmentation results are evaluated by using area error rate and volume calculations and the success rate is calculated as 94.91% over a data set of diverse CTA series of 20 patients according to the evaluation of the expert radiologist. The results show that, the proposed algorithm gives better results especially for atypical liver shapes and low contrast studies where several algorithms fail.

Automatic classification of auditory brainstem responses using SVM-based feature selection algorithm for threshold detection

This paper presents a novel system for automatic recognition of auditory brainstem responses (ABR) to detect hearing threshold. ABR is an important potential signal for determining objective audiograms. Its detection is usually performed by medical experts with often basic signal processing techniques. The proposed system comprises of two stages. In the first stage, for feature extraction, a set of raw amplitude values, a set of discrete cosine transform (DCT) coefficients and a set of discrete wavelet transform (DWT) approximation coefficients are calculated and extracted from signals separately as three different sets of feature vectors. These features are then selected by a modified adaptive method, which mainly supports to the input dimension reduction via selecting the most significant feature components. In the second stage, the feature vectors are classified by a support vector machine (SVM) classifier which is a powerful advanced technique for solving supervised binary classification problem due to its generalization ability. After that the proposed system is applied to real ABR data and it is resulted in a very good sensitivity, specificity and accuracy levels for DCT coefficients such as 99.2%, 94.0% and 96.2%, respectively. Consequently, the proposed system can be used for recognition of ABRs for hearing threshold detection.

Semiautomatic Transfer Function Initialization for Abdominal Visualization Using Self-Generating Hierarchical Radial Basis Function Networks

Being a tool that assigns optical parameters used in interactive visualization, transfer functions (TF) have important effects on the quality of volume rendered medical images. Unfortunately, finding accurate TFs is a tedious and time consuming task because of the trade off between using extensive search spaces and fulfilling the physicians expectations with interactive data exploration tools and interfaces. By addressing this problem, we introduce a semi-automatic method for initial generation of TFs. The proposed method uses a self generating hierarchical radial basis function network to determine the lobes of a volume histogram stack (VHS) which is introduced as a new domain by aligning the histograms of slices of a image series. The new self generating hierarchical design strategy allows the recognition of suppressed lobes corresponding to suppressed tissues and representation of the overlapping regions which are parts of the lobes but can not be represented by the Gaussian bases in VHS. Moreover, approximation with a minimum set of basis functions provides the possibility of selecting and adjusting suitable units to optimize the TF. Applications on different CT/MR data sets show enhanced rendering quality and reduced optimization time in abdominal studies.