Renate Sitte

Comparison of techniques for environmental sound recognition

This paper presents a comprehensive comparative study of artificial neural networks, learning vector quantization and dynamic time warping classification techniques combined with stationary/non-stationary feature extraction for environmental sound recognition. Results show 70% recognition using mel frequency cepstral coefficients or continuous wavelet transform with dynamic time warping.

Neural Networks Approach to the Random Walk Dilemma of Financial Time Series

Predictions of financial time series often show a characteristic one step shift relative to the original data as in a random walk. This has been the cause for opposing views whether such time series do contain information that can be extracted for predictions, or are simply random walks. In this case study, we show that NNs that are capable of extracting weak low frequency periodic signals buried in a strong high frequency signal, consistently predict the next value in the series to be the current value, as in a random walk, when used for one-step-ahead predictions of the detrended S&P 500 time series. In particular for the Time Delay Feed Forward Networks and Elman Networks of various configurations, our study supports the view of the detrended S&P 500 being a random walk series. This is consistent with the long standing hypothesis that some financial time series are random walk series.

Analysis of the predictive ability of time delay neural networks applied to the S&P 500 time series

Reported work on financial time series prediction using neural networks often shows a characteristic one step shift relative to the original data. This seems to imply a failure of the neural network (NN), because a shift corresponds to a random walk prediction. Our systematic analysis of different time delay neural networks predictors applied to the detrended S&P 500 time series, indicates that this prediction behavior is not a limitation of the network, but may be a characteristic of the time series. This suggests that there are no short-term correlations in this stockmarket time series, which is consistent with conventional statistical analysis.

Interactive teaching of elementary digital logic design with WinLogiLab

This paper presents an interactive computerized teaching suite developed for the design of combinatorial and sequential logic circuits. This suite fills a perceived gap in the currently available computer-based teaching software, with the purpose of providing alternative-mode subject delivery. The authors were, therefore, prompted to develop a Microsoft-Windows tutorial suite, WinLogiLab, comprising a set of interactive tutorials that show the link between Boolean algebra and digital combinatorial and sequential circuits. The combinatorial tutorials follow the initial design steps: from Boolean algebra, to truth tables, to minimization techniques, to production of the combinatorial circuit in a seamless way. Similarly, the sequential tutorials can design simple finite-state counters and can model more complex finite-state automata.

Optimizing testing efficiency with error-prone path identification and genetic algorithms

We present a method for optimizing software testing efficiency by identifying the most error prone path clusters in a program. We do this by developing variable length genetic algorithms that optimize and select the software path clusters which are weighted with sources of error indexes. Although various methods have been applied to detecting and reducing errors in a whole system, there is little research into partitioning a system into smaller error prone domains for testing. Exhaustive software testing is rarely possible because it becomes intractable for even medium sized software. Typically only parts of a program can be tested, but these parts are not necessarily the most error prone. Therefore, we are developing a more selective approach to testing by focusing on those parts that are most likely to contain faults, so that the most error prone paths can be tested first. By identifying the most error prone paths, the testing efficiency can be increased.

Recognition of Environmental Sounds Using Speech Recognition Techniques

This paper discusses the use of speech recognition techniques in non-speech sound recognition. It analyses the different techniques used for speech recognition and identifies those that can be used for non-speech sound recognition. It then performs benchmarks on these techniques and determines which technique is better suited for non-speech sound recognition. As a comparison, it also gives results for the use of learning vector quantization (LVQ) and artificial neural network (ANN) techniques in speech recognition.

Layered fluid model and flow simulation for microchannels using electrical networks

Abstract In this paper, we present the modelling of the flowrate of a circular and a rectangular microchannel using an electrical network. The aim of this study is to produce a fast first approximation of the flowrates of microchannels for the design of microfluidic devices. It contributes to the physical component of our virtual reality-prototyping computer-aided design tool for microelectromechanical systems, with emphasis on fast calculations for virtual reality representations. In our model, the flow is segmented into layers and the pertinent models derived. We have achieved this by solving the Navier–Stokes equation, obtaining an analytical model for the circular and a numerical model for the rectangular channels. The resistances of the layers are obtained from the velocity profile of the flow. The electrical network model is implemented in Matlab Simulink. The results are compared with finite element model software (ANSYS) and experimental data.

About the Predictability and Complexity of Complex Systems

With ever-growing complexity of systems to be modeled, there is a strong need for a proper theory of complexity, other than the algorithmic complexity known in computing. The problem is that there is no unanimous consensus as to what complexity is. Several attempts have been made, some are very promising, but a widely applicable theory and practice have not been derived. Quantification is an essential step in modeling to achieve prediction and control of a system. Quantification is also a crucial step in complexity and some complexity quantification models are emerging. In this chapter, a unifying and systematic approach to complexity is proposed. Its aim is to bring some clarity into the unknown, and a step further towards predictability. It serves as an overview and introduction, in particular to the novices on how to deal with complex system as a practical approach. Some practices summarized here are elementary and others are quite ambitious. It happens over and over that the uninitiated researchers make errors, reinvent the wheel or fall into traps. The purpose of this chapter is to offer good advice and a sense on how to avoid pitfalls.

The effect of dynamic design processing for yield enhancement in the fabrication of deep sub-micron MOSFET's

With the downscaling of microelectronic devices, tighter process control and more elaborate fabrication equipment need to be complemented by process correcting techniques if good quality and high yields are to be expected. Dynamic design processing-a forward correcting technique by which some recipe values are recalculated during manufacturing-is such a technique. In this paper the effect of dynamic design processing on deep sub-micron MOSFETs is presented. The results show that a parametric yield improvement in excess of 25% over conventional manufacturing can be achieved. >

Neural Network Systems Technology in the Analysis of Financial Time Series

As human nature strives for wealth and comfort in life, the ways of attaining wealth have changed through time. In the dark ages the focus was on turning material into gold, but nowadays the efforts go into managing and manipulating trades enabling profitable growth in different economic strata. For decades, financial time series predictions have been the target for profitable trade. Those who succeeded are reluctant to share their secrets. Thus successful applications of times series prediction techniques are unlikely to be found in the scholarly literature of the field. We cannot promise a pot of gold, but we will explain financial time series, with emphasis in using neural networks for their prediction, as a technique which has brought significant improvements not only to time series predictions but also enabled new technologies in many other areas. Neural Networks are particularly attractive for the prediction of financial time series because they do not require any kind of model knowledge about the relationships between variables [1]. In this chapter we shall review different approaches to financial time series. We present this material in a way that is understandable to a wide and interdisciplinary audience.