Research Areas
DIAMOND.gif (73 bytes)Artificial Neural Network Modeling
Christos N. Schizas, and Costas Neocleous (Higher Technical Institute, Cyprus), Ibit Esat (Brunel University, UK).

A generic neuronal model using a block form of operational diagrams is investigated. This formalism is used as an aid for classifying neuronal models according to their idiosyncrasies: architecture; activation function and processing element; and the learning rule. The power of this representation is determined by the fact that an easy identification and extraction of the most essential features of single neuron models is allowed. In addition, this representation enables the identification of the truly novel model that have been so far introduced. Furthermore, the generic model facilitates the development of new neuronal structures.

The models are initially presented in a mathematical form, in either continuous or discrete time-space state form, and their performance is analyzed by simulating them using SIMULINK.

The investigated model is believed to be generic enough to be a "parent" of most of the existing models and at the same time detailed enough to appropriately display all important properties of different models introduced so far. For achieving a uniform basis for comparisons, a common notation, based on the control systems representation formalism is used. The adoption of block diagrams employing simple component processes introduces a different understanding of neuronal modeling through graphical visualization. This provides an improved understanding of the structure of a model and thus achieves a deeper insight to the relation between structure and dynamics.


DIAMOND.gif (73 bytes)
Theory and Applications of Genetic Algorithms

Christos N. Schizas, Constantinos S. Pattichis, and K. Christodoulou
(Cyprus Institute of Neurology and Genetics)

The objective of this project is to investigate, how genetics-based machine learning (GBML) can be applied for diagnosing certain neuromuscular disorders based on Electromyographic (EMG) data. The effect of GBML control parameters on diagnostic performance is also examined. A hybrid diagnostic system is introduced that combines features from both neural network and GBML models. Such a hybrid system provides the end-user with a robust and reliable system, since its diagnostic performance relies on more than one learning principle.

In the clinical EMG laboratory, 680 Motor Unit Action Potentials MUAPs) were collected from 12 normal subjects, 11 motor neuron disease, and 11 myopathy subjects. Each subject was described by a 14 component feature vector consisting of the mean and the standard deviation of each of the following MUAP parameters: duration, spike duration; amplitude; area; spike area; phases; and turns. More than a thousand GBML models were developed by varying the following parameters: message length size; number of classifiers; lifetax; period of introducing the genetic algorithm (GA); (expressed in iterations, showing how often the classifier system calls the GA); crossover probability, and mutation probability.

Related Papers

Title :
Creativity in design and artificial neural networks. 
Authors :
Costas Neocleous, Ibit Esat, Christos N. Schizas 
Abstract :
The creativity phase is identified as an integral part of the design phase. The characteristics of creative persons which are relevant to designing artificial neural networks manifesting aspects of creativity, are identified. Based on these identifications, a general framework of artificial neural network characteristics to implement such a goal are proposed.
Presented / Published :
Proc of the Second World Conference “Integrated Design and Process Technology”.
Place / Date :
Austin, Texas. 1996 


Key People
Neocleous C.C (Higher Technical Institute, Cyprus)
Pattichis C. S. (Department of Computer Science,Cyprus)
Schizas C.N. (Department of Computer Science, Cyprus)
Esat I.I.