A computationally efficient neural network-based scheme for wavefront reconstruction and restoration of turbulence-degraded imagery in Adaptive Optics (AO)-based telescopes is described in this paper. Currently popular methods for the estimation of turbulence-generated distortions suffer from high computational complexity that preclude real-time implementations. For overcoming this "curse of dimensionality", a discrete cosine transform (DCT)-based feature extraction scheme that provides a reduced set of features to train a neural network estimator is described in this work. A dimensionality reduction of up to two orders of magnitude, accompanied by a relatively insignificant loss of overall information, and consequently in the overall performance, is achieved by the proposed scheme. Two neural network architectures, Multilayer Perceptron (MLP) and Radial Basis Function (RBF), trained to estimate wavefront parameters are described and their relative performance in AO implementations is outlined. Performance differences measured in terms of specific quantities of interest, such as Strehl ratio, point to the architectural differences and training methods for these to neural networks. The present work represents a novel application of the power of neural networks in facilitating real-time implementation of these systems.

The self-organizing map (SOM) has emerged as one of the popular choices for clustering data; however, when it comes to point density accuracy of codebooks or reliability and interpretability of the map, the SOM leaves much to be desired. In this paper, we compare the newly developed K-means hierarchical (KMH) clustering algorithm to the SOM. We also introduce a new initialization scheme for the K-means that improves codebook placement and, propose a novel visualization scheme that combines the principal component analysis (PCA) and minimal spanning tree (MST) in an arrangement that ensures reliability of the visualization unlike the SOM. A practical application of the algorithm is demonstrated on a challenging bioinformatics problem.

In training a learning machine (LM) with unlimited data samples available in the training set, it is important to be able to determine when the LM has attained an adequate level of generalization in order to stop the training process. While this is a problem that has not yet achieved a satisfactory solution, aiding the determination of the generalization level is the observation that as the LM becomes consistent and reaches an acceptable generalization threshold, finding samples from the training set that would make the system fail and trigger a new cycle of the training algorithm to be implemented becomes more infrequent. In a statistical sense, the number of samples that can be tested as having no new information (i.e. information not already learnt from training cycles already completed) between two successive triggers of training events asymptotically displays a faster than exponential growth behavior, which in turn provides a telltale sign of a LM reaching consistency and thus attaining a desired generalization level. This work employs some ideas taken from statistical learning theory to conjecture the existence of such exponential behavior and designs a new approach to implementing the training steps that can exploit this behavior in order to systematically test the generalization level during the training process. Examples of nonlinear regression problems are included to illustrate the ideas and to validate the methods. The obtained results are general and are independent of the configuration of the LM, its architecture, and the specific training algorithm used; hence, they are applicable to a broad class of supervised learning problems.

Generation of desired trajectory behavior using neural networks involves a particularly challenging spatio-temporal learning problem. This paper introduces a novel solution, i.e., designing a dynamic system whose terminal behavior emulates a prespecified spatio-temporal pattern independently of its initial conditions. The proposed solution uses a dynamic neural network (DNN), a hybrid architecture that employs a recurrent neural network (RNN) in cascade with a nonrecurrent neural network (NRNN). The RNN generates a simple limit cycle, which the NRNN reshapes into the desired trajectory. This architecture is simple to train. A systematic synthesis procedure based on the design of relay control systems is developed for configuring an RNN that can produce a limit cycle of elementary complexity. It is further shown that a cascade arrangement of this RNN and an appropriately trained NRNN can emulate any desired trajectory behavior irrespective of its complexity. An interesting solution to the trajectory modulation problem, i.e., online modulation of the generated trajectories using external inputs, is also presented. Results of several experiments are included to demonstrate the capabilities and performance of the DNN in handling trajectory generation and modulation problems.

Growth in popularity of the Internet has spawned a great interest in collaborative IP networks that support collaborative meetings between individuals or groups located at remote stations. The emphasis on security of information transfer during these meetings has made the management of end-to-end security in collaborative IP network environments, that may involve the creation of ad hoc communication networks that contain the Internet as an intermediate network, an important research issue. Addition of security features through standard methods gives rise to complex incompatibility problems resulting from the specific routing and address translation schemes that may be in place in these networks. The development of enhanced protocols that remove this incompatibility and ensure interoperability between security functions and address translation functions is discussed in this paper. Detailed steps in ensuring end-to-end security in various cases involving change of IP address, change of both IP address and the port, and the use of encapsulation security payload implemented in transport mode are described. The enhanced protocols presented here support generic implementation in the sense that the implementation of the security-related protocol is transparent to the use or not of the address translation scheme. For providing a proof of concept demonstration of the proposed solutions, the structure of a prototype collaborative network, which employs the Internet as an intermediate communication medium for supporting videoconferencing between remote stations, is outlined.

We propose a novel image restoration and super-resolution algorithm based on a convex set theoretic approach. This procedure differs from the currently popular super-resolution algorithms, which typically are derived by employing statistical approaches, in offering a greater flexibility for incorporating scene-related prior information into the restoration processing. Novel methods for modeling and for extrapolation of scene information for efficient use in the restoration process are described. These include a border extraction method that permits a background-foreground separation of the image, a template-based information modeling and a method of extrapolating the image frequency content by a Taylor series expansion. A class of iterative projection algorithms for processing diffraction-limited images by using convex sets modeling various types of a priori information about the object or scene imaged is developed. The restoration and super-resolution performance of these algorithms are established through a variety of simulation experiments.

Designing efficient methods for training dynamic neural networks for learning spatio-temporal patterns is of great interest at present. In particular, the "trajectory generation problem" that involves training the network to learn and replicate autonomously a specified time-varying periodic motion has attracted considerable attention. A systematic approach to solve this problem by decomposing the overall task into two sub-tasks, a spatio-temporal sequence assignment and a mapping of ordered sequences, is presented. This decomposition permits the dynamic neural network to be realized as a cascade of a simple recurrent net followed by a non-recurrent one that yields considerable reduction in training complexity. A detailed performance evaluation of the present scheme is given by considering several trajectory generation experiments that highlight the strong points of this approach, which include simplicity and accuracy in training, flexibility to include control parameters in order to modify online the shape of the trajectory learned and the speed of repetition along a cyclic trajectory, and the possibility of learning both continuous and discontinuous trajectory patterns.

A major problem in a successful deployment of recurrent neural networks in practice is the complexity of training due to the presence of recurrent and feedback connections. The problem is further exacerbated if gradient descent learning algorithms that require computation of error gradients for the necessary updating are used, often forcing one to resort to approximations that may in turn lead to reduced training efficiency. We describe a learning procedure that does not require gradient evaluations and hence offers significant implementation advantages. This procedure exploits the inherent properties of nonlinear simplex optimization in realizing these advantages.

The concept of person-based number is attractive due to several pressing demands in implementation of current wireless communication systems. Some of these are: sudden proliferation in number of area codes; the need for freedom from number changes due to changes in service provider, location, or area code; and demand for improved and integrated communication service for users. Some important factors that determine the efficiency of a mobility management scheme include fewer number of times that locations need to be updated, reduced amount of overhead data that represents mobility related information, and efficient storage mechanisms that allow for fast storage and retrieval of this information. In this paper, we outline a mobility management scheme that provides a globally unique personal number and offers important benefits in regard to these issues. For an efficient implementation of such a scheme however, an investigation of its performance is required in order to estimate signaling traffic and signaling delays. This paper will focus on some recent studies conducted on an analysis of this scheme by using query and update operations as the metrics for determining average call delay and control data storage and transmission requirements. A comparative analysis is also performed with equivalent cases of query and update in the IS-41 system. This analysis provides a valuable tool to determine network management requirements in a person-based number scheme.

Complexity of implementation has been a major difficulty in the development of gradient descent learning algorithms for dynamical neural networks with feedback and recurrent connections. Some insights from the stability properties of the equilibrium points of the network, which suggest an appropriate tailoring of the sigmoidal nonlinear functions, can however be utilized in obtaining simplified learning rules, as demonstrated in this paper. An analytical proof of convergence of the learning scheme under specific conditions is given and some upper bounds on the adaptation parameters for an efficient implementation of the training procedure are developed. The performance features of the learning algorithm are illustrated by applying it to two problems of importance, viz., design of associative memories and nonlinear input-output mapping. For the first application, a systematic procedure is given for training a network to store multiple memory vectors as its stable equilibrium points, whereas for the second application, specific training rules are developed for a three-layer network architecture comprising a dynamical hidden layer for the identification of nonlinear input-output maps. A comparison with the performance of a standard backpropagation network provides an illustration of the capabilities of the present network architecture and the learning algorithm.

Existing methods for image contrast enhancement focus mainly on the properties of the image to be processed while excluding any consideration of the observer characteristics. In several applications, particularly in the medical imaging area, effective contrast enhancement for diagnostic purposes can be achieved by including certain basic human visual properties. Here the authors present a novel adaptive algorithm that tailors the required amount of contrast enhancement based on the local contrast of the image and the observer's Just-Noticeable-Difference (JND). This algorithm always produces adequate contrast in the output image, and results in almost no ringing artifacts even around sharp transition regions, which is often seen in images processed by conventional contrast enhancement techniques. By separating smooth and detail areas of an image and considering the dependence of noise visibility on the spatial activity of the image, the algorithm treats them differently and thus avoids excessive enhancement of noise, which is another common problem for many existing contrast enhancement techniques. The present JND-Guided Adaptive Contrast Enhancement (JGACE) technique is very general and can be applied to a variety of images. In particular, it offers considerable benefits in digital radiography applications where the objective is to increase the diagnostic utility of images. A detailed performance evaluation together with a comparison with the existing techniques is given to demonstrate the strong features of JGACE.

The problem of adaptively managing the transmission bandwidth of an integrated network by implemented optimal partitioning policies for the access control of circuit-switched traffic is studied. Precise formulations within an optimization framework of access control problems, both at an isolated multiplexer level and at the overall network level, are developed. For the simple case of a single multiplexer supporting traffic with identical bandwidth requirements, an iterative scheme with guaranteed fast convergence is presented for the determination of optimal channel allocations. For the more general scenario of a multiplexer supporting traffic with nonidentical bandwidth requirements, a methodology is developed to handle the access control problem by a mixed-integer programming formulation. This methodology is further extended to the network-wide access control scenario, and some implementational issues are discussed. A detailed quantitative evaluation of the proposed algorithm is conducted under different loading conditions, and the corresponding performance and efficiency are compared with those of other known procedures.>

Efficient implementation of a neural network-based strategy for the online adaptive control of complex dynamical systems characterized by an interconnection of several subsystems (possibly nonlinear) centers on the rapidity of the convergence of the training scheme used for learning the system dynamics. For illustration, in order to achieve a satisfactory control of a multijointed robotic manipulator during the execution of high speed trajectory tracking tasks, the highly nonlinear and coupled dynamics together with the variations in the parameters necessitate a fast updating of the control actions. For facilitating this requirement, a multilayer neural network structure that includes dynamical nodes in the hidden layer is proposed, and a supervised learning scheme that employs a simple distributed updating rule is used for the online identification and decentralized adaptive control. Important characteristic features of the resulting control scheme are discussed and a quantitative evaluation of its performance in the above illustrative example is given.

Two enhancements to a recently published hierarchical encryption key management protocol for end-to-end secure communication in internet environments are outlined. The first one concerns a more reliable authentication of the principals which can be realized by a modification of the message structures being exchanged, while the second one concerns a modified protocol that permits the implementation of the hierarchical key management approach in the widely employed TCP/IP-based network interconnections.>

For integrated networks that support circuit-switched traffic classes with varying bandwidth requirements, an efficient integration of the access control and routing functions is important in realizing optimum network performance. A hierarchical scheme for combined access control and routing is presented for realizing the above objective. The hierarchy comprises two levels; at the lower level each network node implements a distributed call routing strategy, while at the higher level a network supervisor implements an access control policy for minimizing the overall blocking probability. The access control problem to be solved at the supervisor level is formulated in an integer programming framework by a partitioning approach which divides the overall network into several distinct subnets, one corresponding to each bandwidth class. For routing the admitted calls within each subnet in an optimal fashion, a distributed minimum loss routing scheme is developed for implementation at the network nodes. Performance evaluation studies showing the superiority of the various control functions developed are presented.>

Several novel results concerning the characterization of the equilibrium conditions of a continuous-time dynamical neural network model and a systematic procedure for synthesizing associative memory networks with nonsymmetrical interconnection matrices are presented. The equilibrium characterization focuses on the exponential stability and instability properties of the network equilibria and on equilibrium confinement, viz., ensuring the uniqueness of an equilibrium in a specific region of the state space. While the equilibrium confinement result involves a simple test, the stability results given obtain explicit estimates of the degree of exponential stability and the regions of attraction of the stable equilibrium points. Using these results as valuable guidelines, a systematic synthesis procedure for constructing a dynamical neural network that stores a given set of vectors as the stable equilibrium points is developed.

Basic results from a qualitative analysis of the exponential stability and equilibrium characterization of a class of dynamical neural networks intended to serve as associative memories are presented. A simple learning rule tailored to efficiently minimize the deviation between the stable equilibrium points of the network and the desired memory vectors to be stored is proposed and is established as a descent procedure for minimizing the deviation. The results are developed for asymmetric interconnection matrices and hence considerably enlarge the scope of the associative memory design compared to existing procedures

A novel multiplexing scheme for integrated networks characterized by the coexistence of circuit-switched and packet-switched traffic is described. The new scheme is realized by reinforcement of the basic movable boundary hybrid-switching technique by incorporating a packet queue-size dependent adaptive TASI (time assigned speech interpolation) mechanism. A detailed queuing model representation of the multiplexer is given, and an analytical evaluation of several performance parameters of interest, such as the blocking probability of the packet-switched subsystem, the freezeout probability of the circuit-switched subsystem, and the expected number of packets in the packet queue, is conducted. The results of some numerical studies are presented to illustrate the multiplexer performance under different conditions. Of particular significance is the tradeoff between the freezeout probability performance of the circuit-switched subsystem and the packet blocking performance of the packet-switched subsystem, which could be exploited in a systematic design of the integrated multiplexing scheme for specific applications.>

The problem of virtual circuit routing in data communication networks under nonstationary conditions is considered. A state variable modeling approach is adopted to describe the dynamic behavior of the possible paths for the virtual circuit through the network. The state model is developed by representing each link in a path as a set of nonlinear differential equations describing the dynamics of the virtual circuit and the network traffic in terms of time-varying mean quantities. An optimal control problem is formulated to determine the virtual circuit routing path which minimizes the number of packets in the network. The solution of the optimal control problem by Hamilton-Jacobi arguments is discussed leading to a threshold routing policy which can be implemented in a decentralized fashion. The performance of the routing algorithm is studied under both steady-state and nonstationary conditions, and its superiority over the performance of previous schemes in a dynamic load environment is demonstrated.>

Numerical techniques for modeling computer networks under nonstationary conditions are discussed, and two distinct approaches are presented. The first approach uses a queuing theory formulation to develop differential equation models which describe the behavior of the network by time-varying probability distributions. In the second approach, a nonlinear differential equation model is developed for representing the dynamics of the network in terms of time-varying mean quantities. This approach allows multiple classes of traffic to be modeled and establishes a framework for the use of optimal control techniques in the design of network control strategies. Numerical techniques for determining the queue behavior as a function of time for both approaches are discussed and their computational advantages are contrasted with simulation.>

A model is presented that precisely describes the mechanism that enforces the security policy and requirements for a multilevel secure network. The mechanism attempts to ensure secure flow of information between entities assigned to different security classes in different computer systems connected to the network. The mechanism also controls the access to the network devices by the subjects (users and processes executed on behalf of the users) with different security clearances. The model integrates the notions of nondiscretionary access control and information flow control to provide a trusted network base that imposes appropriate restrictions on the flow of information among the various devices. Utilizing simple set-theoretic concepts, a procedure is given to verify the security of a network that implements the model.>

Three distinct approaches for modeling computer networks under nonstationary conditions are presented. The first approach uses a queuing-theoretic formulation to develop numerical techniques for determining the nonstationary queue dynamics and for estimating the settling time for the queue. In the second approach, discrete-event simulation techniques are used for the evaluation of network performance during nonstationary periods. The third approach develops a nonlinear state model for representing the dynamics of the packet queues at the various transmission links of the network and establishes a framework for formulating optimal control problems for designing routing and flow control strategies that ensure optimal network performance under both transient and steady-state conditions.>

A hierarchical approach for key management is presented which utilizes the existing network specific protocols at the lower levels and protocols between authentication servers and/or control centers of different networks at the higher levels. Details of this approach are discussed for specific illustrative scenarios to demonstrate the implementation simplicity. A formal verification of the security of the resulting system in the sense of protecting the privacy of privileged information is also conducted by an axiomatic procedure utilizing certain combinatory logic principles. This approach is general and can be used for verifying the security of other existing key management schemes.>

A model which precisely describes the mechanism that enforces the security policy and requirements for a multilevel secure network is described. This mechanism attempts to ensure secure flow of information between entities assigned to different security classes in different computer systems connected to the network. The mechanism also controls the access to the network devices by the subjects (users and processes executed on behalf of the users) with different security clearances. The model integrates the notions of access control and information flow control to provide a trusted network base that imposes appropriate restrictions on the flow of information among the various devices. Utilizing simple set-theoretic concepts, a procedure is given to verify the security of a network that implements the present model.>

A simple analytical approximation to the optimal buffer-management policy valid in the moderate to heavy load ranges is presented. The throughput performance of the present adaptive policies under both steady-state and transient conditions is studied. The superiority of this scheme over the performance resulting from existing buffer-management schemes is demonstrated.>

The problem of designing key management schemes for establishing end-to-end encrypted sessions between source-destinationpairs when the source and the destination are on different networks interconnected via Gateways and intermediate networks is considered. In such an internet environment, the key management problem attains a high degree of complexity due to the differences in the key distribution mechanisms used in the constituent networks and the infeasibility of effecting extensive hardware and software changes to the existing networks. In this paper, a hierarchical approach for key management is presented which utilizes the existing network specific protocols at the lower levels and protocols between Authentication Servers and/or Control Centers of different networks at the higher levels. Details of this approach are discussed for specific illustrative scenarios to demonstrate the implementational simplicity. A formal verification of the security of the resulting system is also conducted by an automatic procedure utilizing certain combinatory logic principles. This approach is general and can be used for verifying the security of any existing key management scheme.

In the implementation of hierarchical schemes for the adaptive control of large communication networks, decomposition of the overall network into node clusters and assignment of a supervisor for each cluster are of central importance. A formulation of the decomposition problem as a mathematical optimization problem is given and an algorithm for solution is presented.

The authors are concerned with the optimization of a class of bilinear systems under an exponential stability constraint by periodic control functions. Such problems arise in many practical applications, a typical one being the design of optimal strategies for cancer chemotherapy by a suitable control of the proliferation kinetics of cell populations. To provide a focus to the approach used for optimization and the utility of periodic controls, this particular application is discussed in detail. An integration of a pharmacokinetic model with a multicompartmental model for cell proliferation is made. This is done in order to obtain a mathematical formulation of the problem of designing treatment strategies as a parameter optimization problem of determining the optimal dose and the optimal period for minimizing the total quantity of drug administered (and hence the host toxicity) under a specified rate of cure constraint. A procedure for solving this problem is developed, and an illustration is made by designing strategies for administering the phase-specific drug ara-c on L1210 leukemia.

A new approach to the development of multilevel control and estimation schemes for large-scale systems with a major emphasis on the reliability of performance under structural perturbations is described. The study is conducted within a decomposition-decentralization framework and leads to simple and noniterative control and estimation schemes. The solution to the control problem involves the design of a set of locally optimal controllers for the individual subsystems in a completely decentralized environment and a global controller on a higher hierarchical level that provides corrective signals to account for the interconnection effects. Similar principles are employed to develop an estimation scheme, which consists of a set of decentralized optimal estimators for the subsystems, together with certain compensating signals for measurements. The principal feature of the paper is a detailed study of the system structure and the consequent classification of interconnection patterns into several interesting categories (beneficial, nonbeneficial, and neutral) based on their effects on decentralized control and estimation.