Hepatocellular carcinoma (HCC) is one of the most critical health problems in the world. For proper treatment, it is important to identify the grade of cancer morbidity from HCC biopsy image. The diagnostic work is not only time-consuming but also subjective. The same biopsy image may be diagnosed as of different grades by different doctors, due to lack of experience or difference in opinion. In this work, we proposed an automatic grading system with classification accuracy matching to an experienced doctor, to help augment the diagnosis process. First, we proposed a segmentation method to isolate all nucleus-like objects present in a biopsy image. Non-target objects (here the target is a single HCC nucleus) present in the biopsy image are isolated too in the segmentation process. To eliminate such non-target objects, we proposed clustering of segmented images and a novel method to filter out target objects. Next, we proposed a two track neural network, where input consists of 2 different images. It combines a single segmented nucleus and a random cropped texture patch of the biopsy image to which the nucleus belongs. At this classifier output, we grade the single nucleus. Finally, a majority voting method is used to identify the grade of the whole biopsy image. We achieved an accuracy of 99.03% for nucleus image grading and 99.66% accuracy for grading biopsy images.

In wireless sensor networks, the energy consumption of sensors is not uniform over the whole region of deployment. The uneven energy usage occurs because some sensors have to transmit data to farther distances or have to transmit more data packets than others. This leads to a shorter duration of operation because some sensors’ energy will deplete fast creating holes in the network. To alleviate this problem, we proposed an algorithm we named Procrastinated Clustering and Multi-Hop Routing (PCMR). To prolong the operation, it will optimally assign sensors with different precomputed procrastination periods to schedule the clustering and routing processes. In PCMR, sensors’ clustering and routing intervals depend on their locations in the network with respect to the sink. The algorithm could reduce and balance energy consumption for sensors distributed over a wide area. Procrastination periods are precomputed off-line before deployment. Therefore, it is easy to implement and is efficient, even for a large network for which real-time reorganization would involve transmitting a large number of signaling packets. The results from simulations show that the proposed PCMR algorithm could balance energy usage among sensors, and prolong the network lifetime compared to existing works based on techniques such as adjusting cluster size and/or multi-path transmission.

In a natural disaster, when there is a sudden rush in vehicle movement from dense residential areas of a city towards some designated evacuation centers, traffic needs to be distributed along the road network in an optimal way so as to minimize the travel time for reaching those evacuation centers. We address this traffic distribution problem in a$k \times k$Manhattan road network with multiple traffic generating sources and a single fixed destination (sink) at a corner point. We first consider queuing delay at each grid point along with a fixed link delay. We then extend our idea to consider a rectangular grid and also real-life road networks where some of the links may be absent. We present a traffic distribution algorithm that would minimize the average time of all vehicles to reach the given destination point. Simulation of our algorithm using SUMO on the grid-structured road map of Manhattan borough in New York city and also on a part of non-grid road map of city of Carbondale, IL USA shows a much shorter average travel time compared to the existing popular algorithms based on dynamic traffic assignments (DTA). The proposed technique is also robust in the sense that even with 20% deviation from optimal assignment of traffic distribution at cross-points, we get significantly better results compared to the existing traffic distribution policies. We then propose a congestion control technique based on detouring of vehicles and show how variable link delay as a function of the volume of traffic flow through a link can be incorporated in our traffic distribution policy. Finally, we show how to extend our technique to consider a scenario with the destination point being anywhere on the grid, instead of being at the corner point.

The efficacy of treatment of cancer depends largely on early detection and correct prognosis. It is more important in case of pulmonary cancer, where the detection is based on identifying malignant nodules in the Computed Tomography (CT) scans of the lung. There are two problems for making correct decision about malignancy: (1) At early stage, the nodule size is small (length 5 to 10 mm). As the CT scan covers a volume of 30cm. × 30cm. × 40cm., manually searching for nodules takes a very long time (approximately 10 minutes for an expert). (2) There are benign nodules and nodules due to other ailments like bronchitis, pneumonia, tuberculosis. To identify whether the nodule is carcinogenic needs long experience and expertise. In recent years, several works have been reported to classify lung cancer using not only the CT scan image, but also other features causing or related to cancer. In all recent works, for CT image analysis, 3-D Convolution Neural Network (CNN) is used to identify cancerous nodules. In spite of various preprocessing used to improve training efficiency, 3-D CNN is extremely slow. The aim of this work is to improve training efficiency by proposing a new deep NN model. It consists of a hierarchical (sliced) structure of recurrent neural network (RNN), where different layers of the hierarchy can be trained simultaneously, decreasing training time. In addition, selective attention (alignment) during training improves convergence rate. The result shows a 3-fold increase in training efficiency, compared to recent state-of-the-art work using 3-D CNN.

We collect environmental information through our sensory organs, perceive them suitably to execute the task in hand. Examples of such tasks are abundant, like driving, operating machines, playing games, hunting, or mushroom picking on mountains. While executing such tasks, there is a marked difference in efficiency and accuracy between an expert and a novice. An expert is attentive to only what is important, still makes fewer errors. She is tacitly aware when or where to focus attention. The operation is efficient because an expert has less information to process, and do not attend to what is irrelevant. An expert correctly perceives from sensory information when to be alert, and therefore she is efficient. At times, the expert is able to explicitly describe her knowledge in a set of rules, but not always. It is not uncommon that the expert herself is unaware how the right decision is taken, and can not express his expertise in explicit rules. This is tacit knowledge acquired through long experience, and lack of which makes a novice ponder to accomplish the task correctly. What is perceived by an expert is different from that of a novice, though the available information through vision, audio and other senses are the same. We can design efficient machines, if the expert's selective attention and perception could be learned and incorporated in machine learning. The motivation of this work is to propose a framework to design machines which will be able to learn the tacit knowledge of an expert. When something important is perceived (like an alarming situation warranting immediate action), it is reflected in bio-signals like increased pulse rate or decrease in GSR. These bio-signals are used as cues to collect labeled data for supervised learning of the tacit knowledge. The system will be efficient by avoiding irrelevant information.

There are various motivations to cluster a set of signals. The common time-series clustering problem is, (1) where the features over the signal time-span of the members, belonging to the same cluster, are same. In other words, the model generating them are same. The other problem could be (2) to divide every member of a multivariate signal into fragments within which individual signals exhibit similar features, which vary from one partition to other. The properties may differ from one signal to other, but the boundaries of the partitions are same. We propose an algorithm to solve problem (2). There are algorithms to divide a single time-series into K segments such that the generating parameters (like regression coefficients) are same for individual segments. It is a complex problem to find the optimum boundaries for K segments and modeling parameters for each segment, minimizing the error. In this work, our target is to segment signals based on features extracted at discrete points in time. Simple clustering algorithms assign same cluster labels based on some similarity measure between feature-vectors, irrespective of their contiguity. As a result, members of a cluster, in general, will be scattered over the time-span of the signal. Members of the same cluster will not be contiguous. To achieve individual clusters to be contiguous, neighboring samples are to be assigned to the same cluster with higher probability, than a sample at further distance, though with similar feature. In our algorithm, similarity measure includes contiguity information of samples from their original physical space. The algorithm consists of calculating the modified similarity metric, and two layers of clustering. The complexity is similar to simple clustering algorithm and therefore it is fast. We applied the algorithm for clustering multivariate EEG signals. The stability of the result over different samples verify the reliability of the algorithm.

Using micro array technology, one can compare gene-expression levels of several thousand genes from two sample cells. Depending on the source of the samples, important investigations, like disease progress, accurate diagnosis, drug response and prognosis after treatment, etc., can be done. The aim of this work is to identify the smallest set of genes, whose expression could classify a target disease with highest accuracy. In this work, we present a two stage feature reduction. In Stage 1, the number of genes are reduced from thousands to N (N ? 100). In Stage 1, individually the relevance of a gene for classification is considered. We proposed a fast algorithm, which can efficiently handle data with more than 2 classes. In Stage 2, we search for the smallest subset of genes, from those selected in Stage 1, so that the classification result is highest. We use wrapper method for feature selection. The selection of the optimum subset of genes is a combinatorial optimization problem, with two optimization criterion - minimizing the cardinality of the subset and maximizing the classification result. Multi-objective optimization Genetic Algorithm is used, and we proposed a new Pareto front which is suitable for the gene-selection problem. The effectiveness of our two stage algorithm is verified using three benchmark gene-expression data, Lung Cancer dataset, SRBC dataset, and MLL dataset.

Feature selection or dimensionality reduction is an important task for any pattern recognition, data mining or machine learning problem. For selection of the optimal subset of relevant features, two steps are needed. In the first step a measure is designed for the evaluation of a candidate feature subset and in the second step, search through the feature space is done for selecting the optimal one. Existing feature selection methodologies use combinations of various evaluation measures and search strategies for selecting optimal feature subset. Though a large number of effective methodologies are already developed, none of them is perfect. Research is still going on to find better algorithm with lesser computational cost. In this work a fuzzy consistency based evaluation measure has been proposed. Consequently a feature selection algorithm using the proposed fuzzy consistency measure with particle swarm optimization, an evolutionary computational technique widely used for optimization problems, is developed for selecting optimal feature subset. Simple simulation experiments with bench mark data sets have been done and the simulation results provide evidence that the proposed algorithm might be a good candidate for selecting optimal feature subset.

In recent years, information on World Wide Web is exploding. Search Engine robots are used to search information on World Wide Web. However, a robot type search engine has a few problems. One problem is that, it is difficult for user to come up with an appropriate query for getting search results she/he intends. Moreover, it is difficult for users to understand contents of search results because a robot type search engine outputs too many search results in a long list format. To resolve these problems, many systems classify a robot type search engine results into clusters. Clusters are labeled and those labels are shown to the user. Cluster labels need to be appropriate words for the web site within the cluster. We have proposed a labeling method using concordance count. First, web search results are obtained by a query input, and the result is classified into clusters. We used our proposed method to assign proper labels to those clusters. To ensure that we use a novel method. We find the set of websites resulted from AND-query using an original query word and the cluster label. If this set and the member of the cluster are common, we say that the concordance count is high. If the concordance count is high, the cluster label is assigned high weight. Finally, we evaluate the accuracy of our proposed method by simulation experiments.

Proton exchange membrane (PEM) fuel cell is one of the promising alternative energy sources with high potential to replace current fossil fuel-based economy. Performance of PEM fuel cells is dependent on several input variables, e.g., electric load, air mass flow, hydrogen mass flow, nitrogen mass flow, cathode water injection rate, etc., in a complex manner. We propose a novel enhanced multilayer perceptron (e-MLP) for modeling the high-power fuel cell stack for estimation of stack voltage and cathode temperature under variation of seven input variables. Essentially, in the proposed e-MLP, pre-processing of input data is first carried out by a nonlinear transformation using orthogonal polynomials. By extensive simulations we have shown that the proposed e-MLP-based model is not only more computationally efficient, but also provides better estimation results in terms of mean square error and correlation coefficient, compared to a recently reported 4-layer MLP-based model.

Failure to deliver intelligent algorithms as was predicted, both by logicist approach of McCarthy and symbolic models of Newell and Simon, as well as successes of connectionist models during 80's, prompted widespread acceptance of artificial neural network algorithms. In particular, multilayer feed forward neural network (aka MLP, an abbreviation from MultiLayer Perceptron) with back-propagation (BP) training was widely accepted as a universal approximator and used for pattern recognition problems, of different complexities. Though MLP takes long training time, once trained it could perform classification/recognition task fast. In this work, we studied how the performance of a trained MLP deteriorates, as its nodes and connection weights are damaged partially. Could the performance be restored, if further training is done on the damaged MLP, the way it was left after damage? How does the restoration depends on the complexity of the task, the network had learned before damage? Is it better to reinitialize the connection weights of the damaged MLP, before starting training again? What are the relative training times and restoration quality? Does it depend on the complexity of the problem? We have investigated these aspects in this work. In addition to explaining our results, we also conjectured to extend possible application of our findings to restore operation of a human brain, damaged by diseases like Alzheimer, or by accident.

Effective and meaningful visualization techniques are quite important for multidimensional DNA microarray gene expression data analysis. Elucidating the cluster properties of these multidimensional data are often complex. Patterns, hypotheses on the relationships, and ultimately of the function of the gene can be analyzed and visualized by non-linear reduction of the multidimensional data to a lower dimension. In this paper, an improved SOM visualization technique named Improved Side Intensity Modulated (ISIM) Self-Organizing Map (SOM) has been proposed and compared with other SOM based visualization techniques. On different datasets, ISIM-SOM is found to offer better cluster boundary, simplicity and clarity.

Delay Tolerant Network (DTN) is a sparse and intermittently connected mobile wireless network where reliable communication and end-to-end connectivity cannot be assured. In DTN, a node delivers messages to the destination using store and forward scheme. Messages are copied and transferred to multiple relay-nodes in order to increase the opportunity for that message to reach the intended destination. Contact is the time duration for which two or more mobile nodes encounter each other within their radio transmission ranges, and are able to transfer messages. Due to arbitrary node movement, contact time will be varied, and is unlikely to be predictable. In this paper, we propose a fuzzy decision mechanism we called ldquofuzzy-sprayrdquo, to improve the routing and packet transfer efficiency in DTN environment. Fuzzy-spray can reduce overall latency in DTN by intelligently selecting appropriate messages to send to the relay-node during its next contact time. It uses fuzzy logic to prioritize messages that are stored in the buffer, based only on local parameters from each message, namely forward transmission count and message size. There is no need to know a priori information about network such as node mobility model, or node-distribution across the deployed area. The simulation results show that fuzzy-spray has the best performance in terms of overall speed of delivery, and lowest average message delay compared with other existing algorithms.

Vehicular ad-hoc network (VANET) is gaining much attention recently because of its many important applications in transportation, to improve road safety, reduce traffic congestion, to enable efficient traffic management etc. However, there are several technical issues to be addressed for its effective deployment. Stability in communication in VANET is difficult to achieve due to rapid network changes. Restoration is inefficient while using traditional protocols based on broadcast storm. In this paper, we propose a new adaptive protocol to improve performance for on road safety alert application in VANET. It can alleviate the broadcast storm problem using adaptive wait-windows and adaptive probability to transmit. Simulation shows that our proposed approach has better performances in terms of number of collision, success rate, and delay, when compared with other existing protocols.

The need for stricter security, availability of sophisticated algorithms and improved hardware at cheaper price, are the driving forces for increasing popularity of biometric authentication. Machine authentication from face images or fingerprints at the entrance of a building or at bank-teller is getting more and more common. The popularity of using face image features for authentication is due to its ease of use. But its success depends on proper orientation and illumination. Facial features change with the angle of orientation, and even a genuine person would be rejected by the machine due to improper orientation of the face towards the camera. In this work, we proposed a multi-layer perceptron based face identification technique which is robust to orientation of the face image under similar illumination condition. Training data of facial features against orientation angle features is used to train a Multi-Layer Perceptron (MLP). It is then used for interpolation of facial features at different angles. Experiments were conducted with two types of face image identifying features, using PCA and ICA. A good interpolation property could be obtained by the trained MLP, and a zero equal error rate (EER) could be achieved.

We propose a novel computationally efficient artificial neural network (NN) for design and development of intelligent sensors to operate in harsh environments which can have wide variation of environmental conditions. The proposed Hermite NN (HeNN) models the inverse characteristics of a sensor and can provide linearized sensor response characteristics irrespective of change in environmental conditions, even when the environmental parameters influence the sensor characteristics nonlinearly. By taking an example of a capacitive pressure sensor, we have shown through extensive computer simulations that the HeNN-based model can linearize its response with maximum full scale (FS) error of plusmn0.5% when it is operated in a harsh environment with temperature variation of -50 to 200degC and influenced nonlinearly. We have compared performance of the proposed HeNN-based model with a MLP-based model and shown its superior performance in terms of FS error and computational complexity.

Increasingly drug discovery is turning to in-silico methods to find potential enzyme or protein-protein inhibitors. The quantitative structure-activity relationship (QSAR) is commonly used to find potential inhibitors in the search for new drugs. In this paper we propose to use a radial basis function (RBF) network to determine the QSAR of aldose reductase inhibitors (ARIs). We find that the RBF network shows promising results of predicting the bioactivities of the ARIs.

In this paper, we present the result of our study on the application of artificial neural networks (ANNs) for adaptive channel equalization in a digital communication system using 4-quadrature amplitude modulation (QAM) signal constellation. We propose a novel single-layer Legendre functional-link ANN (L-FLANN) by using Legendre polynomials to expand the input space into a higher dimension. A performance comparison was carried out with extensive computer simulations between different ANN-based equalizers, such as, radial basis function (RBF), Chebyshev neural network (ChNN) and the proposed L-FLANN along with a linear least mean square (LMS) finite impulse response (FIR) adaptive filter-based equalizer. The performance indicators include the mean square error (MSE), bit error rate (BER), and computational complexities of the different architectures as well as the eye patterns of the various equalizers. It is shown that the L-FLANN exhibited excellent results in terms of the MSE, BER and the computational complexity of the networks.

Different types of sensors are used to control and monitor complex systems in many applications, where the environmental parameters, e.g., temperature, humidity, etc., undergo large variations. In such conditions, the sensor's output may be erroneous and the system being controlled may malfunction. The need of intelligent sensors arise in such situations. These sensors should be capable of compensating for the adverse effects of the environmental conditions on the sensor output and linearization of sensor response, in order to provide correct readout. In this paper, we propose a novel computationally efficient Legendre functional-link artificial neural network (L-FLANN) to develop a smart sensor that can compensate for the adverse effects of the environmental conditions. By taking two types of environmental models and a pressure sensor, we have shown with extensive computer simulations that the proposed smart sensor is computationally efficient with respect to a multi-layer perceptron (MLP)-based sensor model and capable of satisfactory linearization of sensor output. This smart sensor produces only plusmn0.5% full-scale error between the actual and estimated output for the two selected environmental models under a temperature variation of -50 to 200degC.

Genetic algorithm (GA) has some inherent drawbacks which become apparent while trying to solve complex multimodal problems. They are slow and the efficiency depends on parameter values. Some methods were proposed for alleviating these problems. But they did not address all the drawbacks. In this work, we propose a new distributed implementation strategy named variable island GA (VIGA), where the number of islands vary. In VIGA, where the number of individuals in every island is 2, the parameter population size in an island is fixed. Other parameters like number of islands, crossover/mutation probabilities, also need not be set. As the generation progresses, islands are created or erased based on the convergence status of searching in each island. Experiments were done with different function optimization problems. For all experiments VIGA delivered better or at least as good results as obtained by other competitive algorithms, at the expense of less computation and communication costs.

Shotgun sequencing is the state-of-the-art to decode genome sequence. However this technique needs a lot of fragments. Combining those fragments correctly requires enormous computational cost. In our previous work we have shown how genetic algorithm (GA) could solve this problem efficiently. In this work, we added two heuristic ideas with GA to make it more efficient. One is chromosome reduction (CRed) step which shorten the length of the chromosomes, participating in genetic search, to improve the efficiency. The other is chromosome refinement (CRef) step which is a greedy heuristics, rearranging the bits using domain knowledge, to locally improve the fitness of chromosomes. With this hybridization and simple scaffold list, we could obtain longer contigs and scaffolds using GA. We experimented using three actual genome data to test our algorithm. We succeed in restructuring contigs covering about 90% of target genome sequences, and assembling about 500~1,000 fragments into 3 ~ 11 scaffolds. All the experiments were done using common desktop machines.

Mobile host (MH) has to be tracked in personal communication service (PCS) network, for which update and paging signals are required. The number of PCS network subscribers skyrocketed in recent years. To reuse channels over a distance, cell size is reduced and the number of cell crossing by user is becoming high. That makes optimal use of paging and update signal very important. In fact, most MH has unique movement profile, that contains the information of time, route, direction, etc., which is possible to learn and used to predict location. In this paper, we propose mobility pattern based location management scheme using the movement profile. Mobility pattern is learned and system will page only the restricted probable area. We compared the proposed scheme with distance-based location management. Improved cost saving is achieved.

We proposed a genetic algorithm (GA) approach to solve the genome sequencing problem. The main contribution of this work is to add two ideas to improve the efficiency of the algorithm -(1) a chromosome reduction step (CRed) method to shorten the length of the chromosome and thereby the search-space, and (2) chromosome refinement step (CRef) is a greedy heuristics to locally improve the fitness of chromosomes. The algorithm will bring out longer and longer contigs with shorter and shorter gaps, as it continues running. At any stage the user can view the result, stop it when the output serves her/his purpose, or continue for getting longer contigs. We ran the proposed algorithm on part of the Wolbachia project work, and compared the results.

In Bluetooth network environment, device discovery is achieved by the master running the Inquiry protocol and slaves running the Inquiry Scan protocol. The master and a slave establish a connection using the Page and Page Scan protocols, respectively. Once a connection is established between a master and a slave, upper level data can be exchanged between the two devices. Thus, device discovery and connection establishment are fundamental to communication between two Bluetooth devices. While a device is in the process of discovering other devices, there is a random delay. We analyze the discovery time for different situations. We further show the delay-distribution for varied number of devices. We explain reasons of occasional long discovery time, and propose means to improve it.

A new crossover technique, we named rank and proximity based crossover (RPC), to improve the speed and quality of solutions in genetic search, is proposed. In the proposed strategy, the probability of crossover is more when the rank of two chromosomes are both high, and they are closely located in the search space. This probability is again a function of the generation number. In the early stage of genetic search the crossover is independent of the rank and proximity of the partners. Thus, crossover takes place between any two chromosomes selected for crossover, to enable exploration of diverse locations of the problem space. With advancing generations, RPC probabilistically encourages crossover between chromosomes of higher ranks which are closely located in the search space. This ensures avoidance of disruption of good chromosomes by crossover, when prospective locations are found, and thus achieves much faster convergence. Different schemes of this probability function are tried and evaluated, and convergence efficiency is compared with other competitive algorithms.

A distance-based approach is an efficient strategy for location management of mobile hosts in a cellular PCN. But what is the optimum distance for distance-based location management? Does it depend on cell-size, and/or the movement behavior of an individual user, depending on her movement pattern, more efficient? Or should a simple fixed distance for all users be adopted? We have shown that a fixed distance of 4 is nearly optimum for all the mobile hosts, irrespective of a range of variation of their mobility behavior, and the cell size of the network. A more efficient adaptive distance algorithm, where distance is changed according to mobility behavior, is proposed.

The problem of finding an optimum conflict-free transmission schedule for a broadcast packet radio network (PRNET) is NP-complete. In addition to a host of heuristic algorithms, recently, neural network and simulated annealing approaches to solve this problem were reported. We show that the standard genetic algorithm, though able to solve small problems, performed poorly for large networks. This is because classical crossover and mutation operations create invalid members, which flood the whole population, hindering the progress of the search for valid solutions. In this paper, special crossover and mutation operations are defined, such that the members of the population always remain valid solutions of the problem. Excellent results were obtained in a few generations, even for very large networks with 400 nodes. The results were compared with recently reported neural network and mean field annealing approaches.

Since network resources are usually limited, the ability to carry out resource reservation in advance, as well as efficient dynamic multicasting when destination nodes can join and leave the group during the communication period, are essential in all distributed multi-party applications. Total or partial re-routing of a multicast tree for dynamic multicast groups can optimize the tree cost to some extent but they are complex. Furthermore, optimal multicast routing has been proved to be an NP-complete problem. In this paper, we propose an efficient dynamic multicast routing algorithm, which optimizes the total cost of the Steiner tree over the whole session period in an advance resource reservation environment, where information about the resource reservation (i.e. leaving times of participants at the time of joining) are available. The efficiency of our algorithm compared to other existing algorithms is shown by various simulation results.

Unsupervised learning is used to categorize multidimensional data into a number of meaningful classes on the basis of the similarity or correlation between individual samples. In neural-network implementation of various unsupervised algorithms such as principal component analysis (PCA), competitive learning or self-organizing map (SOM), sample vectors are normalized to equal lengths so that similarity could be easily and efficiently obtained by their dot products. In general, sample vectors span the whole multidimensional feature space and existing normalization methods distort the intrinsic patterns present in the sample set. In this work, a novel method of normalization by mapping the samples to a new space of one more dimension has been proposed. The original distribution of the samples in the feature space is shown to be almost preserved in the transformed space. Simple rules are given to map from original space to the normalized space and vice versa.

A new technique called rank based crossover (RBC) to improve the speed of reaching optimal solutions is introduced for genetic algorithms (GAs). In real life, marriages (crossovers) occur between two individuals of similar status in society and/or from neighboring localities. This principle is extended in case of GAs while selecting crossover partners. In the proposed strategy, the probability of crossover is more when their rank in the whole population is close. This probability function changes with advancing generations, so that the effect of RBC is negligible in the beginning and gradually increases. It could easily control fine tuning of the good chromosomes to achieve fast convergence and reach optimum values. Also, the scheme is not centralized like the elitist approach. Different schemes of the probability function are tried and evaluated. The effectiveness of this new method has been demonstrated on the problems of maximizing complex multi-modal functions. The results are compared with standard genetic algorithms (SGA). Another technique called "fitness scaling" is widely used to adaptively scale the objective function to achieve similar goal. We also compared our results with the "linear fitness scaling" strategy. Results using our RBC strategy are found to be superior to those of the fitness scaling method and the SGA in terms of probability of hitting the maximum value as well as speed of finding the maximum.

Routing with multiple QoS constraints is known to be a NP-complete problem. In our previous work (QoSR/sub BF/ algorithm), we have shown that, for a number of specific QoS constraints, this problem can be solved in polynomial time, when a weighted fair queueing (WFQ) service discipline is employed. QoSR/sub BF/ is an on-demand algorithm. We propose a number of QoS routing algorithms for pre-computed paths, when all or some of the flow specification parameters and the amount of bandwidth that has to be reserved are not known a priori. We also show that the size of the routing table could be reduced, as paths satisfying some conditions could be neglected.

In this paper we study QoS based routing algorithm for supporting resource reservation in high-speed Integrated Services Packet Network (ISPN). Recently, this problem was proved to be NP-complete. However, when the considered QoS constraints are bandwidth, delay, delay jitter, and loss free, we have shown that by employing Weighted Fair Queueing (WFQ) service discipline, the complexity of the problem could be reduced to that of shortest path routing without any QoS constraints. Then such a multiple QoS constrained route could be searched in polynomial time. We also present the routing algorithm (called "QoSR/sub BF/"), which is a modified version of Bellman-Ford shortest path algorithm. Simulation results show that QoSR/sub BF/ has better performance compared to the existing QoS routing algorithms.

Much research efforts are presently involved in developing network protocols suitable for multimedia services through LAN and even Internet. Due to large size and time sensitiveness of data, it is usually a hard task to deliver the services up to the expectations of the users. But the basic intention of the user, for the multimedia services, usually have a wide scope for tolerance and flexibility. If that flexibility is explored and exploited, it is possible to basically satisfy the user even with marginal resources. It is important to intelligently interpret user's request as well as distribute the system and network resources optimally. In this paper we are proposing an Intelligent Soft Resource Reservation Protocol (ISORRP) for realizing acceptable running quality of application by intelligent interpretation of users' requests and optimum distribution of available resources among them. In this architecture QoS to a session is only flexibly guaranteed but always to a level acceptable to the user.

Real-time communication networks are designed mainly to support multimedia applications, especially the interactive ones, which require a guarantee of Quality of Service (QoS). Moreover, multicasting is needed as there are usually more than two peers who communicate together using multimedia applications. As for the routing, the network has to find an optimum (least cost) multicast route, that has enough resources to provide or guarantee the required QoS. This problem is called QoS constrained multicast routing and was proved to be an NP-complete problem. In contrast to the existing heuristic approaches, in this paper we propose a modified version of a Hopfield neural network model to solve QoS (delay) constrained multicast routing. By the massive parallel computation of neural networks, it can find a near optimal multicast route very fast, when implemented in hardware. Simulation results show that the proposed model has performance near to the optimal solution and comparable to existing heuristics.

A concept of a flexible network is formalized as a foundation of next generation distributed systems. Considering the requisite properties of flexible networks, we propose an architecture of a flexible network called agent oriented flexible information network (AFIN) model. The model is based on the multi agent paradigm. Furthermore, we focus on a communication network which is a core component of the flexible information network. We propose an agent oriented flexible communication network (AFCN) model and demonstrate with an example the advantages of the proposed architecture of the flexible network.>

As a fundamental study of multiple image processing, the authors have developed a new technique for the analysis of multispectral remote sensing data. Though the basis of the authors' algorithm is similar to a histogram analysis, they have proposed a novel way of the representation of multidimensional image data, which facilitates a nonexpert to locate and assign different classes present in the data set. The authors' experimentation was done with NOAA satellite image data. The brightness of the received data from channels of different frequency bands are the different dimensions of the multispectral image. By using this method, the authors could successfully classify NOAA satellite data received from the northern part of Japan, and locate the plane areas as an exercise. The authors have quantitatively compared their result with official data. The authors' result was found to be only 1 percent in deviation with the official data.>