This paper presents a novel approach to dynamic transmission bandwidth allocation for transport of real-time variable-bit-rate video in ATM networks. Video traffic statistics are measured in the frequency domain: the low frequency signal captures the slow time-variation of consecutive scene changes while the high frequency signal exhibits the feature of strong frame autocorrelation. Our queueing study indicates that the video transmission bandwidth in a finite-buffer system is essentially characterized by the low frequency signal. We further observe in typical JPEG/MPEG video sequences that the time scale of video scene changes is in the range of a second or longer, which localizes the low frequency video signal in a well-defined low frequency band. Hence, in a network design it is feasible to implement dynamic allocation of video transmission bandwidth using on-line observation and prediction of scene changes. Two prediction schemes are examined: recursive least square method and time...

Neural networks have been shown to be a promising tool for forecasting financial time series. Several design factors significantly impact the accuracy of neural network forecasts. These factors include selection of input variables, architecture of the network, and quantity of training data. The questions of input variable selection and system architecture design have been widely researched, but the corresponding question of how much information to use in producing high-quality neural network models has not been adequately addressed. In this paper, the effects of different sizes of training sample sets on forecasting currency exchange rates are examined. It is shown that those neural networks---given an appropriate amount of historical knowledge ---can forecast future currency exchange rates with 60 percent accuracy, while those neural networks trained on a larger training set have a worse forecasting performance. In addition to higher-quality forecasts, the reduced training set sizes reduce development cost and time.

Artificial neural networks have been used to support applications across a variety of business and scientific disciplines during the past years. Artificial neural network applications are frequently viewed as black boxes which mystically determine complex patterns in data. Contrary to this popular view, neural network designers typically perform extensive knowledge engineering and incorporate a significant amount of domain knowledge into artificial neural networks. This paper details heuristics that utilize domain knowledge to produce an artificial neural network with optimal output performance. The effect of using the heuristics on neural network performance is illustrated by examining several applied artificial neural network systems. Identification of an optimal performance artificial neural network requires that a full factorial design with respect to the quantity of input nodes, hidden nodes, hidden layers, and learning algorithm be performed. The heuristic methods discussed in this paper produce optimal or near-optimal performance artificial neural networks using only a fraction of the time needed for a full factorial design. Keywords: Artificial neural networks; Heuristics; Input vector; Hidden layer size; ANN learning method; Design. Heuristics Principles for the Design of Artificial Neural Networks - Page 3 1.

Keywords: Tra#c-Monitoring, Number-plate Recognition, Vehicle Tracking, Computer Vision, Character Recognition, Object Recognition.

In this paper we compare the ability of a fuzzy neural network and a classical backpropagation network to classify odour samples which were obtained by an electronic nose employing semi-conducting oxide conductometric gas sensors. Two different samples sets were analysed: first the aroma of 3 blends of commercial coffee, and secondly the headspace of 6 different tainted water samples. The two experimental data-sets provided an excellent opportunity to test the ability of a fuzzy neural network due to the high level of sensor variability often experienced with this type of sensor. Results are presented on the application of 3 layer fuzzy neural networks to electronic nose data which demonstrate a considerable improvement in performance to a common back-propagation network. 1. Introduction Artificial neural networks (ANNs) have been the subject of considerable research for over twenty years. However, it is during the last decade or so that research interest 1 Sensors and Actuators, vo...

Two ANNA neural-network chips are integrated on a 6U VME board, to serve as a high-speed platform for a wide variety of algorithms used in neural-network applications as well as in image analysis. The system can implement neural networks of variable sizes and architectures, but can also be used for #ltering and feature extraction tasks that are based on convolutions. The board contains a controller implemented with #eld programmable gate arrays #FPGA#, memory and bus interfaces, all designed to support the high compute power of the ANNA chips. Compared to a previous evaluation board #1#, this new system is designed for maximum speed and is roughly 10 times faster than the previous board. The system has been tested for such tasks as text location, character recognition, and noise removal as well as for emulating cellular neural networks #CNN#. A sustained speed of up to 2 billion connections per second #GC#s# and a recognition speed of 1000 characters per second has been measured. Keyw...

Introduction In the previous chapter we reviewed the different ANN paradigms, their learning characteristics etc. Since feedforward neural networks with back-propagation learning is the most widely used configuration, we deal with this paradigm in this part of the book in more detail. Since the backpropagation learning for a bigger network with a large training set takes a long time ( in terms of days), it becomes imperative to look at parallel implementation schemes to reduce this large training time. Before a detailed study of this can be undertaken, it is worthwhile to survey the different parallel implementation schemes for BP networks which exist in the literature indicating their strengths and weaknesses. This is the main theme of this chapter. Since the implementations depend very much on the type of parallelism used, the type of processor networks and also the topology, a brief description and definition of these terms are given in the beginning of the chapter. After th

Search engines perform the task of retrieving information related to the user-supplied query words. This task has two parts; one is finding ’featured words ’ which describe an article best and the other is finding a match among these words to user-defined search terms. There are two main independent approaches to achieve this task. The first one, using the concepts of semantics, has been implemented partially. For more details see another paper of Marko et al., 2002. The second approach is reported in this paper. It is a theoretical model based on using Neural Network (NN). Instead of using keywords or reading from the first few lines from papers/articles, the present model gives emphasis on extracting ’featured words ’ from an article. Obviously we propose to exclude prepositions, articles and so on, that is, English words like "of, the, are, so, therefore, " etc. from such a list. A neural model is taken with its nodes pre-assigned energies. Whenever a match is found with featured words and userdefined search words, the node is fired and jumps to a higher energy. This firing continues until the model attains a steady energy level and total energy is now calculated. Clearly, higher match will generate higher energy; so on the basis of total energy, a ranking is done to the article indicating degree of relevance to the user’s interest. Another important feature of the proposed model is incorporating a semantic module to refine the search words; like finding association among search words, etc. In this manner, information retrieval can be improved markedly. 1.

Two NET3K neural-network chips are integrated on a board system with an SBus interface, to serve as a high speed image analysis platform. The system is optimized for convolutional networks. Up to 6 Kernels of size 16 x 16 pizels are scanned simulta- neously over an image. In this way, simple geometric shapes are eztracted from an image, representing its content in a compact form. A standard processor can then do the high level interpretation. To pre- vent I/O bottlenecks between board and host several high speed programmable logic devices are included on the board, to implement tapped delay lines and com- pression/decompression algorithms. The board can process 20 frames per second, achieving over 100 GC/s (billion connections per second). The SBus interface makes it possible to directly "plug" the board into a SUN Sparcstation, providing a compact and low cost solution for cornpies image analysis tasks. Several document processing applications are described.

Since the earliest days of the electronic computer, there has always been a small group of people who have seen the