Neural and wavelet network models for financial distress classification

This work analyzes the use of linear discriminant models, multi-layer perceptron neural networks and wavelet networks for corporate financial distress prediction. Although simple and easy to interpret, linear models require statistical assumptions that may be unrealistic. Neural networks are able to discriminate patterns that are not linearly separable, but the large number of parameters involved in a neural model often causes generalization problems. Wavelet networks are classification models that implement nonlinear discriminant surfaces as the superposition of dilated and translated versions of a single "mother wavelet" function. In this paper, an algorithm is proposed to select dilation and translation parameters that yield a wavelet network classifier with good parsimony characteristics. The models are compared in a case study involving failed and continuing British firms in the period 1997-2000. Problems associated with over-parameterized neural networks are illustrated and the Optimal Brain Damage pruning technique is employed to obtain a parsimonious neural model. The results, supported by a re-sampling study, show that both neural and wavelet networks may be a valid alternative to classical linear discriminant models.

Comparison of extrasystolic ECG signal classifiers using discrete wavelet transforms

This work compares and contrasts results of classifying time-domain ECG signals with pathological conditions taken from the MITBIH arrhythmia database. Linear discriminant analysis and a multi-layer perceptron were used as classifiers. The neural network was trained by two different methods, namely back-propagation and a genetic algorithm. Converting the time-domain signal into the wavelet domain reduced the dimensionality of the problem at least 10-fold. This was achieved using wavelets from the db6 family as well as using adaptive wavelets generated using two different strategies. The wavelet transforms used in this study were limited to two decomposition levels. A neural network with evolved weights proved to be the best classifier with a maximum of 99.6% accuracy when optimised wavelet-transform ECG data wits presented to its input and 95.9% accuracy when the signals presented to its input were decomposed using db6 wavelets. The linear discriminant analysis achieved a maximum classification accuracy of 95.7% when presented with optimised and 95.5% with db6 wavelet coefficients. It is shown that the much simpler signal representation of a few wavelet coefficients obtained through an optimised discrete wavelet transform facilitates the classification of non-stationary time-variant signals task considerably. In addition, the results indicate that wavelet optimisation may improve the classification ability of a neural network. (c) 2005 Elsevier B.V. All rights reserved.

Cooled infrared filters and dichroics for the James Webb Space Telescope Mid-Infrared Instrument

The cooled infrared filters and dichroic beam splitters manufactured for the Mid-Infrared Instrument are key optical components for the selection and isolation of wavelengths in the study of astrophysical properties of stars, galaxies, and other planetary objects. We describe the spectral design and manufacture of the precision cooled filter coatings for the spectrometer (7 K) and imager (9 K). Details of the design methods used to achieve the spectral requirements, selection of thin film materials, deposition technique, and testing are presented together with the optical layout of the instrument. (C) 2008 Optical Society of America.

High-performance infrared narrow-bandpass filters for the Indian National Satellite System meteorological instrument (INSAT-3D)

This paper describes the design and manufacture of a set of precision cooled (210K) narrow-bandpass filters for the infrared imager and sounder on the Indian Space Research Organisation (ISRO) INSAT-3D meteorological satellite. We discuss the basis for the choice of multilayer coating designs and materials for 21 differing filter channels, together with their temperature-dependence, thin film deposition technologies, substrate metrology, and environmental durability performance. (C) 2008 Optical Society of America.

Synthesis method for visible and infrared broadband spaceflight antireflection coatings

Evolutionary synthesis methods, as originally described by Dobrowolski, have been shown in previous literature to be an effective method of obtaining anti-reflection coating designs. To make this method even more effective, the combination of a good starting design, the best suited thin-film materials, a realistic optimization target function and a non-gradient optimization method are used in an algorithm written for a PC. Several broadband anti-reflection designs obtained by this new design method are given as examples of its usefulness.

Improvements in or relating to the manufacture of multilayer interference filters

This invention relates to the manufacture of multi-layer interference filters for use with infra-red radiation, especially at wavelengths beyond 3.8 microns. A method of manufacturing a multi-layer interference filter comprising the steps of forming on a substrate successive layers of lead telluride and another material in alternation, under conditions in which sufficient oxygen is included in the lead telluride layers to reduce the apparent free charge carrier concentration therein, so that the resulting filter exhibits enhanced transparency to radiation of wavelengths greater than 3.8 microns and enhanced natural absorption to radiation of wavelength less than 3.8 microns.

Neural networks for linear control: an analysis

This paper considers the use of radial basis function and multi-layer perceptron networks for linear or linearizable, adaptive feedback control schemes in a discrete-time environment. A close look is taken at the model structure selected and the extent of the resulting parameterization. A comparison is made with standard, nonneural network algorithms, e.g. self-tuning control.

A critique of neural networks for discrete-time linear control

This paper discusses the use of multi-layer perceptron networks for linear or linearizable, adaptive feedback.control schemes in a discrete-time environment. A close look is taken at the model structure selected and the extent of the resulting parametrization. A comparison is made with standard, non-perceptron algorithms, e.g. self-tuning control, and it is shown how gross over-parametrization can occur in the neural network case. Because of the resultant heavy computational burden and poor controller convergence, a strong case is made against the use of neural networks for discrete-time linear control.

Spectral characterisation of cooled filters for remote sensing

The improvements obtained on cooling atmospheric remote-sensing instruments for space flight applications has promoted research in characterization of the necessary optical filters. By modelling the effects of temperature on the dispersive spectrum of some constituent thin film materials, the cooled performance can be simulated and compared. multilayer filter designs with the measured spectra from actual filters. Two actual filters are discussed, for the 7µm region, one a composite cut-on/cut-off design of 13% HBW and the other an integral narrowband design of 4% HBW.

Systems analysis of absorbing media

A complete treatment of the charcterastic admittance-matrix representation of multilayer thin-film systems with absorbing media is presented. The algorithm from the systems analysis is implemented on an IBM microcomputer and some examples of filter design calculation are presented. Relevant source code in IBM Advanced Basic interpreter are also included.

Measurement of infrared multilayer filters at temperatures down to 4 deg K

Measurement is reported at 4 deg K (and blocked transmission below 10-5) of PbTe/ZnS thin-film filters deposited on Ge substrates. The reduced carrier-absorption which is obtained by cooling these PbTe films is found to accord with simple theory. Advantage for various high-performance multilayers by cooling is significant at the longer wavelengths, and has been verified.

Durability assessment of PbTe/II-IV infrared filters (Space Shuttle 1st LDEF)

Residual stress having been further reduced, selected infrared coatings composed of thin films of (PbTe/ ZnS (or ZnSe) can now be made which comply with the durability requirements of MIL-48616 whilst retaining transparency. Such improved durability is due to the sequence:- i) controlled deposition, followed by ii) immediate exposure to air, followed by iii) annealing in vacuo to relieve stress. (At the time of writing we assume the empiric procedure "exposure to air/annealing in vacuo" acts to relieve the inherent stresses of deposition). As part of their testing, representative sample filters prepared by the procedure are being assembled for the shuttle's 1st Long Duration Exposure Facility (to be placed in earth orbit for a considerable period and then recovered for analysis). The sample filters comprise various narrowband-designs to permit deduction of the constituent thin film optical properties. The Reading assembly also contains representative sample of the infrared crystals, glasses, thin-film absorbers and bulk absorbers, and samples of shorter-wavelength filters prepared similarly but made with Ge/SiO. Findings on durability and transparency after exposure will be reported.

Digital signal processing of optical monitoring by means of a Kalman filter

A Kalman filter algorithm has been applied to interpret the optical reflectance excursions during vacuum deposition of infrared coatings and multilayer thin-film filters. The application has been described in detail elsewhere and this paper now reports on-line experience for estimating deposition rate and thickness. The estimation proved sufficiently reliable to firstly 'navigate' regular manufacture (as controlled by a skilled operator) and to subsequently reproduce the skill without interpretation or intervention whilst maintaining exemplary product quality. Optical control by means of this Kalman filter application is therefore considered suitable as a basis for the automated manufacture of infrared coatings and multilayer thin-film filters.

Modelling the propagation of terahertz radiation through a tissue simulating phantom

Terahertz (THz) frequency radiation, 0.1 THz to 20 THz, is being investigated for biomedical imaging applications following the introduction of pulsed THz sources that produce picosecond pulses and function at room temperature. Owing to the broadband nature of the radiation, spectral and temporal information is available from radiation that has interacted with a sample; this information is exploited in the development of biomedical imaging tools and sensors. In this work, models to aid interpretation of broadband THz spectra were developed and evaluated. THz radiation lies on the boundary between regions best considered using a deterministic electromagnetic approach and those better analysed using a stochastic approach incorporating quantum mechanical effects, so two computational models to simulate the propagation of THz radiation in an absorbing medium were compared. The first was a thin film analysis and the second a stochastic Monte Carlo model. The Cole–Cole model was used to predict the variation with frequency of the physical properties of the sample and scattering was neglected. The two models were compared with measurements from a highly absorbing water-based phantom. The Monte Carlo model gave a prediction closer to experiment over 0.1 to 3 THz. Knowledge of the frequency-dependent physical properties, including the scattering characteristics, of the absorbing media is necessary. The thin film model is computationally simple to implement but is restricted by the geometry of the sample it can describe. The Monte Carlo framework, despite being initially more complex, provides greater flexibility to investigate more complicated sample geometries.

Optimising memory usage in n-tuple neural networks

The use of n-tuple or weightless neural networks as pattern recognition devices is well known (Aleksander and Stonham, 1979). They have some significant advantages over the more common and biologically plausible networks, such as multi-layer perceptrons; for example, n-tuple networks have been used for a variety of tasks, the most popular being real-time pattern recognition, and they can be implemented easily in hardware as they use standard random access memories. In operation, a series of images of an object are shown to the network, each being processed suitably and effectively stored in a memory called a discriminator. Then, when another image is shown to the system, it is processed in a similar manner and the system reports whether it recognises the image; is the image sufficiently similar to one already taught? If the system is to be able to recognise and discriminate between m-objects, then it must contain m-discriminators. This can require a great deal of memory. This paper describes various ways in which memory requirements can be reduced, including a novel method for multiple discriminator n-tuple networks used for pattern recognition. By using this method, the memory normally required to handle m-objects can be used to recognise and discriminate between 2^m — 2 objects.