Khalid Al-Mashouq

The use of neural nets to combine equalization with decoding for severe intersymbol interference channels

This paper deals with the problem of combining equalization with decoding in channels which have severe intersymbol interference. A multilayer neural net structure is proposed to achieve the process of equalization and decoding simultaneously. Experimental examples show that this method results in a substantial improvement over the more conventional methods of performing equalization and decoding.

The use of neural nets to combine equalization with decoding

Conventionally, equalization and decoding are performed independently in cascade. A multilayer neural network is proposed to perform these two tasks simultaneously. The proposed method is compared with the conventional methods in the case of severe intersymbol interference (ISI). With a proper structure and training, the neural-net equalizer/decoder is shown to outperform substantially the more conventional decoders.<<ETX>>

Minimizing memory loss in learning a new environment

Abstract Human and other living species can learn new concepts without losing the old ones. On the other hand, artificial neural networks tend to “forget” old concepts. In this paper, we present three methods to minimize the loss of the old information. These methods are analyzed and compared for the linear model. In particular, a method called network sampling is shown to be optimal under certain condition on the sampled data distribution. We also show how to apply these methods in the nonlinear models.

Clustered Recursive Branching Network

Recursive branching network (RBN) was proposed in [1] to solve linearly non-separable problems using output-coded perceptrons. It relies on splitting the training patterns, at random, between parallel perceptrons. However, the random splitting mechanism can trap the perceptron with conflicting patterns. Optimizing the splitting methods, through clustering, is proposed here to ensure meaningful way of distributing the patterns between the perceptrons. We propose four splitting methods which use different similarity measures between patterns. We examine these methods on five standard data sets. In general, these methods enhance the performance of RBN and in many cases contribute to reducing the network complexity compared with random-splitting RBN.

Prediction of Neural Net Tolerance to Noise

To protect a feed-forward neural net against errors, error correcting codes have been used in previous studies to encode the output label. Here we analyze the effect of additive noise on the performance of a 1-layer coded net and compare it to an uncoded net. The derived results are then used to predict the performance of any multilayer neural net.

Call Quality and Its Parameter Measurement in Telecommunication Networks

The measurement of call quality from end users perspective is emerging area of research on speech quality in telecommunication networks. Our idea is focused at deriving and developing a system to measure certain call parameters during the call and provide average scores at the end of the call. Call quality for the bundle of calls is derived based on the aggregation of successful call parameters which gives the overall call quality measure. The call parameters used in our research were Signal Strength, the successful call rate, normal drop call rate, handover drop rate. GPS coordinates are also used to locate the location and quality of the individual calls. Also a model using the sms feature for tackling the critical quality has been proposed and implemented. Finally the basic bandwidth quality measurement approach is presented which addresses the issue of low bandwidth quality with respect to both user and the operator.

Synchronization problem in replicated HLR — HLR audit tool

GSM (global system for mobile communication) uses a centralized, standalone network database referred to as HLR (home location register). Each of the HLRs occurs in pairs (primary and mated). This paper presents a solution to the problem of mismatches within a pair of HLRpsilas that can happen due to non perfect operations, such as provisioning process carried out by GSM operators. An efficient solution to find mismatches and rectify them is presented. The performance of the system has been investigated on a GSM operator network with few pairs of HLRs, with more than 2 million subscribers in each HLR pairs. HLR pair records with mismatching parameters were provided and corrected with satisfactory results.

Optimization of recursive branching network

Recursive branching network (RBN) was proposed by Al-Mashouq (1997) to solve linearly non-separable problems using output-coded perceptrons. It relies on splitting the training patterns, at random, between parallel perceptrons. However, the random splitting mechanism can trap the perceptron in conflicting patterns. Optimized splitting methods are proposed here to ensure meaningful way of splitting. We propose three splitting methods which use different similarity measures between patterns. We examine these methods on five standard data sets. In general, these methods enhance the performance of RBN and in many cases contribute to lowering the network complexity.

Recursive branching network

The capacity of a 1-layer net is limited compared to a multilayer net. However, there is no explicit rule for optimal structuring and training of a multilayer net. Thus iterative methods are usually used. Here we propose a systematic way to build and train a special multilayer network called recursive branching network. The theory behind this network is presented along with experimental work done on VOWEL data set.