Search PubMed⌕ Search

Biomedical subjects

K Khorasani

Publications and source records attributed to K Khorasani.

5 recordsLinked to original sources
PubMed2005

Constructive feedforward neural networks using hermite polynomial activation functions.

In this paper, a constructive one-hidden-layer network is introduced where each hidden unit employs a polynomial function for its activation function that is different from other units. Specifically, both a structure level as well as a function level adaptation methodologies are utilized in constructing the network. The functional level adaptation scheme ensures that the "growing" or constructive network has different activation functions for each neuron such that the network may be able to capture the underlying input-output map more effectively. The activation functions considered consist of orthonormal Hermite polynomials. It is shown through extensive simulations that the proposed network yields improved performance when compared to networks having identical sigmoidal activation functions.

Algorithms↗
PubMed2004

New training strategies for constructive neural networks with application to regression problems.

Regression problem is an important application area for neural networks (NNs). Among a large number of existing NN architectures, the feedforward NN (FNN) paradigm is one of the most widely used structures. Although one-hidden-layer feedforward neural networks (OHL-FNNs) have simple structures, they possess interesting representational and learning capabilities. In this paper, we are interested particularly in incremental constructive training of OHL-FNNs. In the proposed incremental constructive training schemes for an OHL-FNN, input-side training and output-side training may be separated in order to reduce the training time. A new technique is proposed to scale the error signal during the constructive learning process to improve the input-side training efficiency and to obtain better generalization performance. Two pruning methods for removing the input-side redundant connections have also been applied. Numerical simulations demonstrate the potential and advantages of the proposed strategies when compared to other existing techniques in the literature.

Algorithms↗
PubMed1998

Neural network based control schemes for flexible-link manipulators: simulations and experiments.

This paper presents simulation and experimental results on the performance of neural network-based controllers for tip position tracking of flexible-link manipulators. The controllers are designed by utilizing the modified output re-definition approach. The modified output re-definition approach requires only a priori knowledge about the linear model of the system and no a priori knowledge about the payload mass. Four different neural network schemes are proposed. The first two schemes are developed by using a modified version of the 'feedback-error-learning' approach to learn the inverse dynamics of the flexible manipulator. Both schemes require only a linear model of the system for defining the new outputs and for designing conventional PD-type controllers. This assumption is relaxed in the third and fourth schemes. In the third scheme, the controller is designed based on tracking the hub position while controlling the elastic deflection at the tip. In the fourth scheme which employs two neural networks, the first network (referred to as the 'output neural network') is responsible for specifying an appropriate output for ensuring minimum phase behavior of the system. The second neural network is responsible for implementing an inverse dynamics controller. The performance of the four proposed neural network controllers is illustrated by simulation results for a two-link planar flexible manipulator and by experimental results for a single flexible-link test-bed. The networks are all trained and employed as online controllers and no off-line training is required.

Journal Article↗
PubMed1996

An Adaptive Structure Neural Networks with Application to EEG Automatic Seizure Detection.

This paper introduces a new algorithm for adaptively adjusting the structure of a multi-layer back-propagation network. The proposed algorithm belongs to the class of neuron generating strategies as opposed to the class of neuron pruning strategies. Initially a "small" multi-layer perceptron network is selected. The stabilized error is used as an index to determine whether the network needs to generate a new neuron or not. If after a period of learning the error is stabilized, but the error is larger than a desired value, then new neuron(s) is (are) generated. The new neurons are placed at locations that contribute most to the network error behavior through the fluctuation in their input weight vectors. Among the features of the new architecture are its improved performance and generalization capabilities compared to a standard fixed-structure back-propagation network. Application to an electroencephalogram (EEG) automatic epileptic seizure detection is presented to illustrate advantages and capabilities of the proposed algorithm. Using an actual data from five patients it is shown that the proposed approach correctly identifies all true seizures that are also identified by an expert physician. The new algorithm provides a reduction of 60-70% in the training epochs as compared to a back-propagation algorithm. Furthermore, it is shown that by utilizing a new training algorithm it is possible to reduce the false seizure detections to zero while resulting in a 5.1% error in identifying the true seizures. Copyright 1996 Elsevier Science Ltd

Journal Article↗