Snjezana Soltic

KNOWLEDGE EXTRACTION FROM EVOLVING SPIKING NEURAL NETWORKS WITH RANK ORDER POPULATION CODING

This paper demonstrates how knowledge can be extracted from evolving spiking neural networks with rank order population coding. Knowledge discovery is a very important feature of intelligent systems. Yet, a disproportionally small amount of research is centered on the issue of knowledge extraction from spiking neural networks which are considered to be the third generation of artificial neural networks. The lack of knowledge representation compatibility is becoming a major detriment to end users of these networks. We show that a high-level knowledge can be obtained from evolving spiking neural networks. More specifically, we propose a method for fuzzy rule extraction from an evolving spiking network with rank order population coding. The proposed method was used for knowledge discovery on two benchmark taste recognition problems where the knowledge learnt by an evolving spiking neural network was extracted in the form of zero-order Takagi-Sugeno fuzzy IF-THEN rules.

Evolving spiking neural networks for taste recognition

The paper investigates the use of the spiking neural networks for taste recognition in a simple artificial gustatory model. We present an approach based on simple integrate-and-fire neurons with rank order coded inputs where the network is built by an evolving learning algorithm. Further, we investigate how the information encoding in a population of neurons influences the performance of the networks. The approach is tested on two real-world datasets where the effectiveness of the population coding and networkpsilas adaptive properties are explored.

Towards the Optimum Light Source Spectrum

This paper is concerned with designing light source spectra for optimum luminous efficacy and colour rendering. We demonstrate that it is possible to design light sources that can provide both good colour rendering and high luminous efficacy by combining the outputs of a number of narrowband spectral constituents. Also, the achievable results depend on the numbers and wavelengths of the different spectral bands utilized in the mixture. Practical realization of these concepts has been demonstrated in this pilot study which combines a number of simulations with tests using real LEDs (light emitting diodes). Such sources are capable of providing highly efficient lighting systems with good energy conservation potential. Further research is underway to investigate the practicalities of our proposals in relation to large-scale light source production.

FPGA implementation of an evolving spiking neural network

This research presents a Field Programmable Gate Array (FPGA) implementation of a taste recognition model. The model is based on simple integrate and fire neurons and facilitates an on-line learning. The whole system, including the hardware required to build (evolve) the network was hosted on one FPGA chip. The implementation used 45% of the logic elements, 76% of the memory, and 23% of the dedicated multiplier slices of the chip. FPGA size was sufficient for 64 neurons with up to 64 synapses each (a total of 4096 synapses). The proposed FPGA implementation was successfully applied to a classification problem of taste recognition and the FPGA implementation was at least 10 times faster when evolving the network and 74 times faster during the classification than the software simulations executed by a stand-alone PC.

Optimization of laser-based white light illuminants.

The authors build on previous experience in the optimization of white-light sources based on combinations of narrow-band spectra. They extend those concepts by using delta-function spectra to study the prospects of future optimal laser-based sources. The optimization process is based on a trade-off between the color rendering properties and the luminous efficacy of the radiation. Optimal solutions for four, five and six delta-function spectra with correlated color temperatures in the 3000 to 5500 K range are presented and analyzed. White-light sources with these properties would likely find wide acceptance in numerous lighting applications.

Light source optimization: spectral design and simulation of four-band white-light sources

The authors build on previous experience in the optimization of white light sources based on combinations of known narrow-band spectra. The authors extend these concepts by using synthetic (mathematically generated) four-band spectra to study the prospects for the spectral design of future optimal light sources. The optimization is based on achieving a trade-off of the color rendering properties against the luminous efficacy of the radiation, using the differential evolution (DE) algorithm. It is shown that this trade-off is controllable through the choice of the optimization parameters, specifically the choice of terms in the optimization fitness function. Depending on the choice of the shapes for the synthetic spectral bands, the results are comparable with those achieved in earlier work on the optimization of four-band light-emitting diode (LED) spectra. However, the range of color temperatures achieved in the present study is considerably wider than was previously found for optimum four-band LED mixtures.

Ecological Informatics for the Prediction and Management of Invasive Species

Ecologists face rapidly accumulating environmental data form spatial studies and from large-scale field experiments such that many now specialize in information technology. Those scientists carry out interdisciplinary research in what is known as ecological informatics. Ecological informatics is defined as a discipline that brings together ecology and computer science to solve problems using biologically-inspired computation, information processing, and other computer science disciplines such as data management and visualization. Scientists working in the discipline have research interests that include ecological knowledge discovery, clustering, and forecasting, and simulation of ecological dynamics by individual-based or agent-based models, as well as hybrid models and artificial life. In this chapter, ecological informatics techniques are applied to answer questions about alien invasive species, in particular, species that pose a biosecurity threat in a terrestrial ecological setting. Biosecurity is defined as the protection of a regionʼs environment, flora and fauna, marine life, indigenous resources, and human and animal health. Because biological organisms can cause billions of dollars of impact in any country, good science, systems, and protocols that underpin a regulatory biosecurity system are required in order to facilitate international trade. The tools and techniques discussed in this chapter are designed to be used in a risk analysis procedure so that agencies in charge of biosecurity can prioritize scarce resources and effort and be better prepared to prevent unexpected incursions of dangerous invasive species. The methods are used to predict, (1) which species out of the many thousands might establish in a new area, (2) where those species might establish, and, (3) where they might spread over a realistic landscape so that their impact can be determined.

Dynamic Neuro-fuzzy Inference and Statistical Models for Risk Analysis of Pest Insect Establishment

The paper introduces a statistical model and a DENFIS-based model for estimating the potential establishment of a pest insect. They have a common probability evaluation module, but very different clustering and regression modules. The statistical model uses a typical K-means algorithm for data clustering, and a multivariate linear regression to build the estimation function, while the DENFIS-based model uses an evolving clustering method (ECM) and a dynamic evolving neural-fuzzy inference system (DENFIS) respectively. The predictions from these two models were evaluated on the meteorological data compiled from 454 worldwide locations, and the comparative analysis shows advantages of the DENFIS-based model as used for estimating the potential establishment of a pest insect.

Influence of Peak Wavelengths on Properties of Mixed-LED White-Light Sources

The purpose of this investigation is to quantify the influence of the peak wavelength shifts in commercially available LEDs on the characteristics of the mixed-LED white-light sources. For this purpose, a tetrachromatic spectrum was optimized and then subjected to deviations in the peak wavelengths. A total of 882 combinations of peak wavelength values were evaluated, and the results are reported in terms of correlated colour temperature, colour-rendering properties, and radiant luminous efficacy. The results show that there can be significant changes in the characteristics of the source under these conditions. Such changes are highly likely to present problems when dealing with applications where an effective and accurate white-light source is important.

Application of the CIECAM02 colour appearance model to predict the effect of gamma on the colours viewed on CRT monitors

Abstract The purpose of this investigation is to develop a prediction process for the subjective response of human viewers to the colours of a range of samples displayed on a CRT (cathode ray tube) monitor, for a range of different values of monitor gamma. We have used the gain-offset-gamma (GOG) model of the behaviour of the CRT, together with the assumption of “standard” CRT colorimetry, to permit an assessment of the influence of monitor gamma on the appearance of the colours reproduced on its screen. Colour differences have been computed in a colour space, based on the CIECAM02 colour appearance model, here termed the (J,aC,bC) colour space. A total of 76 numerically defined colour samples were used, and were subjected to the influence of eight different gamma values. The results show that there can be significant colour appearance differences as a consequence of erroneous CRT gamma values. Such errors in colour appearance are highly likely to present problems when dealing with applications where an accurate impression or understanding the displayed colour is important.