The time allocation problem in multi-function cognitive radar systems focuses on the trade-off between scanning for newly emerging targets and tracking the previously detected targets. We formulate this as a multi-objective optimization problem and employ deep reinforcement learning to find Pareto-optimal solutions and compare deep deterministic policy gradient (DDPG) and soft actor-critic (SAC) algorithms. Our results demonstrate the effectiveness of both algorithms in adapting to various scenarios, with SAC showing improved stability and sample efficiency compared to DDPG. We further employ the NSGA-II algorithm to estimate an upper bound on the Pareto front of the considered problem. This work contributes to the development of more efficient and adaptive cognitive radar systems capable of balancing multiple competing objectives in dynamic environments.

In this paper, scanning and multi-target tracking in a radar system are considered, and adaptive radar resource management is analyzed. In particular, time management in radar scanning and tracking of multiple maneuvering targets subject to budget constraints is studied with the goal to jointly maximize the tracking and scanning performances of a cognitive radar. The constrained optimization of the dwell time allocation to each target is addressed via deep Q-network (DQN) based reinforcement learning. In the proposed constrained deep reinforcement learning (CDRL) algorithm, both the parameters of the DQN and the dual variable are learned simultaneously. Numerical results show that radar can autonomously allocate more time to the tracking task that requires greater attention while providing time for scanning and also constraining the total time budget below the predefined threshold.

The dynamics of charitable donor co-attendance networks can help fundraisers assess and improve fundraising outcomes. To improve understanding of donor-giving patterns, this study examines a large, multi-year network describing the co-attendance of donors at charitable fundraising events. We analyze the dynamics of co-attendance networks based on their topological structure, shift in node characteristics, and various network properties. Among other results, we observe a 76% increase in giving value for donors that showed increased centrality rank over nonoverlapped snapshots. In the data we examined, 19.14% of the donors whose giving increased and 16.24% of donors that remained in the same giving range exhibited increased co-attendance with high-capacity donors, whereas none of the donors that shifted to a lower class exhibited increased co-attendance with high-capacity donors over the periods, potentially illustrating a positive peer effect on donors. Some similarity was also observed in the giving characteristics of donors who co-attend events, with a 0.211 assortativity coefficient for the giving class of donors as a characteristic of donors when considering network dynamics using a rolling window size of 3 years. This is followed by analyzing the group-level similarities that reveal an interlinked clique of communities with diverse sizes. Our results show that large communities have a higher fraction of wealthy donors.

We tackle the challenge of Zero-Shot identification of authors of source code, which can be used with no prior samples of authors outside of the training data. In our approach, a feedforward neural network is first trained on a multi-class classification task. Then, a substantial part of this network is duplicated and reused to compare code samples. We refer to this design as Feedforward Duplicated Resolver (FDR) model. We propose new input features to train this model, called Variable-Independent Nested Bigrams, extracted from the Abstract Syntax Trees of code samples. These features provide robustness against lexical and layout obfuscation attacks frequently used in plagiarism attempts. This approach performs accurately even on code samples from unknown authors, on data obtained from Google Code Jam, an international coding competition platform. For example, for the task of predicting whether a pair of samples from 43 unknown authors have been written by the same person, we obtain an AUC of 0.96 and 0.91 for non-obfuscated and obfuscated code, respectively.

Gene Regulatory Networks (GRNs) represent the functional interactions between genes, such as the expression of one gene enhancing the subsequent expression of another gene. GRNs can be inferred from observations of the expressions of various genes over time. This task is difficult, due to the complex interactions among collections of genes, as well as the inherent noise in the observable data. In realistic genomes with thousands of genes, millions of gene-pairs exist, although the actual number of possible interactions may be considerably smaller, since each gene may affect only a small number of other genes. This motivates the application of sparse learning algorithms for GRN inference. In particular, we explore evolutionary algorithms, and their applications with sparse matrix representations. Our approach can speed up the optimization process and find good solutions, uncovering the underlying GRNs. We study evolution strategies, particle swarm optimization, and a greedy algorithm, and compare their performance in terms of solution quality and computational effort required for GRN inference.

The smart grid interconnects a power grid (network) and a communication network, and enables bi-directional flow of electricity and information. To prevent the cascading failures which occur when the disruptions in one network cause disruptions in the other network, robustness should be enhanced by increasing the number of links (edges) between the power grid and the information flow network. Given a budget which constrains the number of new links that can be added to `strengthen' the network, the best strategy to determine where to add those new links remains an open research problem. This paper presents a multi-objective approach to identify the best locations in the power network where new links can be added, to improve the overall robustness of the smart grid when constrained by resource limitations. Simulation results show that substantially greater robustness is obtained by using this approach, when compared to other link addition algorithms.

In social networks that change with time, an important problem is the prediction of new links that may be formed in the future. Existing works on link prediction have focused only on networks where links are permanent, an assumption that is not valid in many real world social networks. In many real world networks, in addition to new links being created, existing links also get removed. In this paper, we extend existing link prediction methods and apply a supervised learning algorithm to networks with non-permanent links. The results we obtain on Twitter @-mention networks show that our method performs very well in such networks.

A rising star is an individual who shows the potential to become a star in the near future. We investigate the problem of finding rising stars when heterogeneous data sources are available to define the same person. The proposed solution examines multiple data sources to determine how the importance of an individual improves over time. Scores from different data sources are combined using a multi-objective optimization approach, as well as a rank aggregation approach. Compared with existing methods, our approach identifies rising stars with higher h-index, number of papers and citation count in the academic domain. Further, we show that the new approach can also be applied to other data sources such as the Information Security Stack Exchange question answer forum.

In this paper, we address the constrained resource allocation problems arising in the context of spectrum sharing in cognitive radio networks utilizing a multi-dimensional formulation. Given the activity of the primary users (PUs), we consider multiple objectives and constraints, viz., sum rate, fairness, number of active secondary users (SUs), power consumption, and quality of service requirements (of both PUs and SUs). The three dimensions for the optimization task are the assignment of power, frequency, and antenna directionality to various SUs. Efficient heuristic algorithms are developed for five variations of the NP-hard optimization problems. Solution quality tradeoffs are shown for three algorithms, viz., convex relaxation with tree pruning, convex relaxation with gradual removal, and a genetic algorithm (GA); results show that the GA provides a reasonable balance between solution quality and computational effort. The multi-objective problems are solved using a modification of the NSGA-II evolutionary algorithm, obtaining a set of Pareto-optimal solutions under computational constraints.

Identification of a set of key players in a given social network is of interest in many disciplines such as sociology, politics, finance, and economics. Each of the current algorithms for this task addresses a single objective, but does not perform well from the perspective of other objectives. In real life applications, we need a set of key players which can perform well with respect to multiple objectives of interest. In this paper, we propose a new perspective for key player identification, based on optimizing multiple objectives of interest, and illustrate its applicability. In addition we propose an algorithm to select the most suitable sets of key players when the user can identify a subset of objectives as important. We apply these algorithms to the Eventual Influence Limitation problem and show that our multi-objective approach outperforms previous approaches.

In most social networks that are observed over time, we find that some individuals leave and others join the network. It is often important to modify the connections in the resulting network to satisfy desired properties associated with the network as well as individual nodes. We formulate this as a multi-objective optimization problem that requires maximization of two measures: the network Information Flow Quality (IFQ) and the Personal Satisfaction Quality(PSQ). Algorithms are developed to accomplish these optimization tasks, and shown to result in satisfactory network reconfiguration.

Gene regulatory networks (GRNs) describe epistatic relationship of genes and how the expression of some genes influence the expression of other genes. This information is critical for understanding molecular mechanisms regulating various biological processes and molecular basis of several diseases. Current research work mostly attempts to infer such regulatory relationships (and GRN architecture topology) from gene expression data. This paper improves on this methodology by utilizing additional information available that describes which cis-elements are present in which genes, and at what locations. Using the underlying principle that target genes of a transcription factor should share the same binding site in their promoter regions, we propose a scoring method that facilitates the refinement of a candidate GRN. Improvements are demonstrated with three data sets, on which GRNs are first obtained from existing dataset (AtRegNet) or using an existing approach (ARACNe), and then modified using cis-element information.

Abstract Wireless sensor networks can be employed to provide distributed real‐time navigation instructions to users attempting to travel in hazardous environments. In this work, we propose a distributed path planning algorithm for sensor network navigation in dynamic hazardous environments. Using geographic or virtual coordinates of sensors and based on a partial reversal method for directed acyclic graphs (DAG), our algorithm constructs a distributed in‐network directed navigation graph, where each source sensor is guaranteed to have at least one desired directed path to one destination sensor. When the hazardous environment changes due to its dynamic nature, path replanning does not need to reconfigure most of the directed links in the graph unaffected by the changes. Correctness of our algorithm is proved and extensive simulation results demonstrate that the constructed navigation graph provides near‐optimal navigation paths for users, successfully adapts to dynamic hazardous environments, and requires very low communication overhead for maintenance, when compared to other navigation graphs constructed by existing algorithms that use frequent or periodic flooding. Copyright © 2010 John Wiley & Sons, Ltd.

For distributed detection in a wireless sensor network, sensors arrive at decisions about a specific event that are then sent to a central fusion center that makes global inference about the event. For such systems, the determination of the decision thresholds for local sensors is an essential task. In this paper, we study the distributed detection problem and evaluate the sensor thresholds by formulating and solving a multiobjective optimization problem, where the objectives are to minimize the probability of error and the total energy consumption of the network. The problem is investigated and solved for two types of fusion schemes: 1) parallel decision fusion and 2) serial decision fusion. The Pareto optimal solutions are obtained using two different multiobjective optimization techniques. The normal boundary intersection (NBI) method converts the multiobjective problem into a number of single objective-constrained subproblems, where each subproblem can be solved with appropriate optimization methods and nondominating sorting genetic algorithm-II (NSGA-II), which is a multiobjective evolutionary algorithm. In our simulations, NBI yielded better and evenly distributed Pareto optimal solutions in a shorter time as compared with NSGA-II. The simulation results show that, instead of only minimizing the probability of error, multiobjective optimization provides a number of design alternatives, which achieve significant energy savings at the cost of slightly increasing the best achievable decision error probability. The simulation results also show that the parallel fusion model achieves better error probability, but the serial fusion model is more efficient in terms of energy consumption.

We propose a set of distributed algorithms for in-network path planning that enables a distributed sensor network navigation service in dynamic environments. Different from existing algorithms that use frequent or periodic flooding, our algorithms exploit geographic information of sensors to construct and maintain navigation links. Based on a partial reversal method of directed acyclic graphs, our algorithms ensure that each source sensor has at least one safe navigation path to one of the multiple destination sensors.

A new algorithm, HAMMER, discovers cis-elements in promoter regions of the co-regulated genes. We show that HAMMER is faster and more accurate than well-known tools currently in use to identify cis-elements. Given input sequences that represent promoter regions of genes, this algorithm searches for subsequences of desired length w whose frequency of occurrence is relatively high, while accounting for slightly corrupted variants (with up to d substitutions). Various w-mers are numerically encoded and represented in a hash table, and d-neighbors are efficiently discovered using a modulo-4 arithmetic operation. Profile matrices are constructed and evaluated using a high-order Markov model based on background data (from a gene database). HAMMER discovers the most frequently occurring w-mers (permitting corruption in at most d positions). Experiment results show that HAMMER is significantly faster and discovers more motifs present in the test sequences, when compared with two well-known motif-discovery tools (MDScan and AlignACE).

Efficient localization of a ballistic missile is an important task in missile defense problems. This paper formulates and solves the sensor placement problem for efficient estimation of the missile location. The first part of this paper develops a mathematical framework for the localization of the missile using multiple sensors based on Cramer-Rao lower bound (CRLB) analysis. We derive the Fisher information matrix to facilitate the evaluation of estimation accuracy. The second part of the paper presents an evolutionary algorithm for obtaining the sensor placements. Simulation results show that the evolutionary algorithm outperforms a greedy sensor placement algorithm and obtains sensor placements with very low estimation error.

This paper considers the problem tracking a moving target in a multisensor environment using distributed particle filters (DPFs). Particle filters have a great potential for solving highly nonlinear and non-Gaussian estimation problems, in which the traditional Kalman filter (KF) and extended Kalman filter (EKF) generally fail. How ever, in a sensor network, the implementation of distributed particle filters requires huge communications between local sensor nodes and the fusion center. To make the DPF approach feasible for real time processing and to reduce communication requirements, we approximate a posteriori distribution obtained from the local particle filters by a Gaussian mixture model (GMM). We propose a modified EM algorithm to estimate the parameters of GMMs obtained locally. These parameters are transmitted to the fusion center where the best linear unbiased estimator (BLUE) is used for fusion. Simulation results are presented to illustrate the performance of the proposed algorithm

This paper presents an information theoretic perspective on design and analysis of evolutionary algorithms. Indicators of solution quality are developed and applied not only to individuals but also to ensembles, thereby ensuring information diversity. Price's Theorem is extended to show how joint indicators can drive reproductive sampling rate of potential parental pairings. Heritability of mutual information is identified as a key issue.

Commonly used fitness measures, such as mean squared error, often fail to reward individuals whose presence in the population is necessary to explain substantial portions of the data variance. Diversity indicators are often arbitrary, may reflect diversity irrelevant to solving the problem, and are incommensurate with fitness measures. By contrast, information theoretic functional are computable general indicators of fitness and diversity without these typical failings. We propose normalized mutual information, redundancy and synergy measures for genetic programming. We also propose selection for recombination and survival by "pairing potential" and "pair potential" estimation, and offer numerical examples as empirical support for theoretical claims.

An approach for data fusion in wireless sensor networks involves the use of mobile agents that selectively visit the sensors and incrementally fuse the data, thereby eliminating the unnecessary transmission of irrelevant or non-critical data. The order of sensors visited along the route determines the quality of the fused data and the communication cost. We model the mobile agent routing problem as a multi-objective optimization problem, maximizing the total detected signal energy while minimizing the energy consumption and path loss. Simulation results show that this problem can be solved successfully using evolutionary multi-objective algorithms such as EMOCA and NSGA-II. This approach also enables choosing between two alternative routing algorithms, to determine which one results in higher detection accuracy.

The problem of distributed sequential detection in the presence of communication constraints is considered. The observations available at each sensor are first compressed to multibit sensor decisions and sent to the fusion center. At the fusion center, a sequential data fusion scheme is implemented in order to reach a global decision. An algorithm is developed for optimal bandwidth distribution among sensors under a fixed bandwidth constraint. Under symmetry and conditional independence assumptions, the algorithm can be simplified substantially: the cooperative quantizer design algorithm reduces to independent quantizer design. The case when communication bandwidth is the only constraint is also considered.

In many problems, the quality of solutions and computational effort required by optimization algorithms can be improved by exploiting knowledge found in the linkages or interrelations between problem dimensions or components. These linkages are sometimes known a priori from the nature of the itself; in other cases linkages can be learned by sampling the data space prior to the application of the optimization algorithm. This paper presents a new version of the particle swarm optimization algorithm (PSO) that utilizes linkages between components, performing more frequent simultaneous updates on subsets of particle position components that are strongly linked. Prior to application of this linkage-sensitive PSO algorithm, problem specific linkages can be learned by examining a randomly chosen collection of points in the search space to determine the correlations in fitness changes resulting from perturbations in pairs of components of particle positions. The resulting algorithm, adaptive-linkage PSO (ALiPSO) has performed significantly better than the classical PSO, in simulations conducted so far on several test problems.

Fragment assembly is an important task in genome sequencing, and involves reconstructing a long sequence based on many overlapping subsequences (fragments) that contain errors. This paper presents a fragment assembler using a new parallel hierarchical adaptive variation of evolutionary algorithms. The innovative features include a new measure for evaluating sequence assembly quality and the development of a hybrid algorithm. Results from the simulation of 56 cases demonstrate that sequence assembly by the new sequencing method is highly accurate and noise-tolerant.

This paper presents a modification of the particle swarm optimization algorithm (PSO) intended to combat the problem of premature convergence observed in many applications of PSO. The proposed new algorithm moves particles towards nearby particles of higher fitness, instead of attracting each particle towards just the best position discovered so far by any particle. This is accomplished by using the ratio of the relative fitness and the distance of other particles to determine the direction in which each component of the particle position needs to be changed. The resulting algorithm (FDR-PSO) is shown to perform significantly better than the original PSO algorithm and some of its variants, on many different benchmark optimization problems. Empirical examination of the evolution of the particles demonstrates that the convergence of the algorithm does not occur at an early phase of particle evolution, unlike PSO. Avoiding premature convergence allows FDR-PSO to continue search for global optima in difficult multimodal optimization problems.

Multi-phase particle swarm optimization is a new algorithm to be used for discrete and continuous problems. In this algorithm, different groups of particles have trajectories that proceed with differing goals in different phases of the algorithm. On several benchmark problems, the algorithm outperforms standard particle swarm optimization, genetic algorithm, and evolution programming.

We have developed a tool that facilitates dynamically updating beliefs with time. This tool addresses directed probabilistic inference networks that may contain cycles, and takes into account the time delays associated with observations and decisions. Relevance of different observers may decay at different rates in the same application, and the belief in a hypothesis decays towards the associated prior probability. Simple models with few parameters have been implemented, with a user interface that facilitates changes to the structure and parameters of the graphical model, and associated conditional probabilities.

The vehicle routing problem with time windows (VRPTW) is a very important problem in the transportation industry since it occurs frequently in everyday practice, e.g. in scheduling bank deliveries. Many heuristic algorithms have been proposed for this NP-hard problem. This paper reports the successful application of GrEVeRT (Graph-based Evolutionary algorithm for the Vehicle Routing Problem with Time windows), an evolutionary algorithm based on a directed acyclic graph model. On well-known benchmark instances of the VRPTW, we obtain better results than those reported by other researchers using genetic algorithms.

Problem-specific knowledge is often implemented in search algorithms using heuristics to determine which search paths are to be explored at any given instant. As in other search methods, utilizing this knowledge will more quickly lead a genetic algorithm (GA) towards better results. In many problems, crucial knowledge is not found in individual components, but in the interrelations between those components. For such problems, we develop an interrelation (linkage) based crossover operator that has the advantage of liberating GAs from the constraints imposed by the fixed representations generally chosen for problems. The strength of linkages between components of a chromosomal structure can be explicitly represented in a linkage matrix and used in the reproduction step to generate new individuals. For some problems, such a linkage matrix is known a priori from the nature of the problem. In other cases, the linkage matrix may be learned by successive minor adaptations during the execution of the evolutionary algorithm. This paper demonstrates the success of such an approach for several problems.

Fine-honing the crossover operator to produce higher fitness children is shown to result in improved genetic search. To illustrate this, two new general-purpose crossover operators are described. These operators require more computation time than traditional crossover operators, but the number of fitness evaluations and the overall amount of time spent by the genetic algorithm (to obtain solutions of desired near-optimal quality) is reduced significantly.

A new optimization method has been proposed by J. Kennedy and R.C. Eberhart (1997; 1995), called Particle Swarm Optimization (PSO). This approach combines social psychology principles and evolutionary computation. It has been applied successfully to nonlinear function optimization and neural network training. Preliminary formal analyses showed that a particle in a simple one-dimensional PSO system follows a path defined by a sinusoidal wave, randomly deciding on both its amplitude and frequency (Y. Shi and R. Eberhart, 1998). The paper takes the next step, generalizing to obtain closed form equations for trajectories of particles in a multi-dimensional search space.

Crew scheduling is an NP-hard constrained combinatorial optimization problem, which is very important for the airline industry. We propose a genetic algorithm, GraGA, to solve this problem. A new graph based representation utilizes memory effectively, and provides a framework in which we can easily develop various genetic operators.

Efficient new algorithms are presented for maximum-likelihood and suboptimal soft-decision decoding algorithms for linear block codes. The first algorithm, MA*, improves the efficiency of the A* decoding algorithm, conducting the heuristic search through a code tree while exploiting code-specific properties. The second algorithm, H*, reduces search space by successively estimating the cost of the minimum-cost codeword with a fixed value at each of the most reliable and linearly independent components of the received message. The third algorithm, directed search, finds the codeword closest to the received vector by exploring a continuous search space. The strengths of these three algorithms are combined in a hybrid algorithm, applied to the (128,64), the (256,131), and the (256,139) binary-extended Bose-Chaudhuri-Hocquenghem (BCH) codes. Simulation results show that this hybrid algorithm can efficiently decode the (128,64) code for any signal-to-noise ratio, with near-optimal performance. Previously, no practical decoder could have decoded this code with such a performance for all ranges of signal-to-noise ratio.

Shape recognition is a challenging task when images contain overlapping, noisy, occluded, partial shapes. This paper addresses the task of matching input shapes with model shapes described in terms of features such as line segments and angles. The quality of matching is gauged using a measure derived from attributed shape grammars. We apply genetic algorithms to the partial shape-matching task. Preliminary results, using model shapes with 6 to 70 features each, are extremely encouraging.

We study two classes of sigmoids: the simple sigmoids, defined to be odd, asymptotically bounded, completely monotone functions in one variable, and the hyperbolic sigmoids, a proper subset of simple sigmoids and a natural generalization of the hyperbolic tangent. We obtain a complete characterization for the inverses of hyperbolic sigmoids using Euler's incomplete beta functions, and describe composition rules that illustrate how such functions may be synthesized from others. These results are applied to two problems. First we show that with respect to simple sigmoids the continuous Cohen-Grossberg-Hopfield model can be reduced to the (associated) Legendre differential equations. Second, we show that the effect of using simple sigmoids as node transfer functions in a one-hidden layer feedforward network with one summing output may be interpreted as representing the output function as a Fourier series sine transform evaluated at the hidden layer node inputs, thus extending and complementing earlier results in this area. Copyright 1996 Elsevier Science Ltd

A framework has been developed for an expert system that generates process plans for the manufacture of machined parts. The system consults the solid models of the parts (CAD database) for geometrical and technological information, and generates a process plan. Process plans are generated using the blackboard architecture concept through interaction among different experts from the process planning domain. A systematic method has been proposed for creating the different knowledge and databases as required for process planning. The methodology also provides an efficient scheme for exploiting the process knowledge bases. A prototype expert system (BBPP) has been developed using the CLIPS V 6.0 expert system shell and has been integrated with a feature based design system developed over the CAEDS solid modeler.

The rate of convergence of net output error is very low when training feedforward neural networks for multiclass problems using the backpropagation algorithm. While backpropagation will reduce the Euclidean distance between the actual and desired output vectors, the differences between some of the components of these vectors increase in the first iteration. Furthermore, the magnitudes of subsequent weight changes in each iteration are very small, so that many iterations are required to compensate for the increased error in some components in the initial iterations. Our approach is to use a modular network architecture, reducing a K-class problem to a set of K two-class problems, with a separately trained network for each of the simpler problems. Speedups of one order of magnitude have been obtained experimentally, and in some cases convergence was possible using the modular approach but not using a nonmodular network.

Partitioning graphs into equally large groups of nodes, minimizing the number of edges between different groups, is an extremely important problem in parallel computing. This paper presents genetic algorithms for suboptimal graph partitioning, with new crossover operators (KNUX, DKNUX) that lead to orders of magnitude improvement over traditional genetic operators in solution quality and speed. Our method can improve on good solutions previously obtained by using other algorithms or graph theoretic heuristics in, minimizing the total communication cost or the worst case cost of communication for a single processor. We also extend our algorithm to incremental graph partitioning problems, in which the graph structure or system properties changes with time.>

Abstract The paper presents a new approach that uses neural networks to predict the performance of a number of dynamic decentralized load‐balancing strategies. A distributed multicomputer system using distributed load‐balancing strategies is represented by a unified analytical queuing model. A large simulation data set is used to train a neural network using the back‐propagation learning algorithm based on gradient descent The performance model using the predicted data from the neural network produces the average response time of various load balancing algorithms under various system parameters. The validation and comparison with simulation data show that the neural network is very effective in predicting the performance of dynamic load‐balancing algorithms. Our work leads to interesting techniques for designing load balancing schemes (for large distributed systems) that are computationally very expensive to simulate. One of the important findings is that performance is affected least by the number of nodes, and most by the number of links at each node in a large distributed system.

Soft-decision decoding is an NP-hard problem of great interest to developers of communication systems. We show that this problem is equivalent to the problem of optimizing Walsh polynomials. We present genetic algorithms for soft-decision decoding of binary linear block codes and compare the performance with various other decoding algorithms including the currently developed A* algorithm. Simulation results show that our algorithms achieve bit-error-probabilities as low as 0.00183 for a [104,52] code with a low signal-to-noise ratio of 2.5 dB, exploring only 22,400 codewords, whereas the search space contains 4.5 × 10l5 codewords. We define a new crossover operator that exploits domain-specific information and compare it with uniform and two-point crossover.

We present an encoding algorithm for lattices that significantly reduces space requirements while allowing fast computations of least upper bounds and greatest lower bounds of pairs of elements. We analyze the algorithms for encoding, LUB and GLB computations, and prove their correctness. Empirical experiments reveal that our method is significantly more space efficient than the transitive closure method, and the saving becomes increasingly more important as the size of the lattice increases.>

The problem of analyzing images containing multiple sparse overlapped patterns is addressed. This problem arises naturally when analyzing the composition of organic macromolecules using data gathered from their NMR spectra. Using a neural network approach, excellent results are obtained in using NMR data to analyze the presence of various amino acids in protein molecules. High correct classification percentages (about 87%) are achieved for images containing as many as five substantially distorted overlapping patterns.

The backpropagation algorithm converges very slowly for two-class problems in which most of the exemplars belong to one dominant class. An analysis shows that this occurs because the computed net error gradient vector is dominated by the bigger class so much that the net error for the exemplars in the smaller class increases significantly in the initial iteration. The subsequent rate of convergence of the net error is very low. A modified technique for calculating a direction in weight-space which decreases the error for each class is presented. Using this algorithm, the rate of learning for two-class classification problems is accelerated by an order of magnitude.

The relationship between the number of hidden nodes in a neural network, the complexity of a multiclass discrimination problem, and the number of samples needed for effect learning are discussed. Bounds for the number of samples needed for effect learning are given. It is shown that Omega(min (d,n) M) boundary samples are required for successful classification of M clusters of samples using a two-hidden-layer neural network with d-dimensional inputs and n nodes in the first hidden layer.

The relationship between psychological performance and psychological data for human auditory frequency discrimination is addressed. Two theories had been proposed earlier to explain the data available for this task: the single-band theory and the multiband theory. By constructing neural network models of the physiological structure, the authors attempted to substantiate these theories. However, the predictions of both these theories do not correspond well with observed data. The authors formulate a new weighted-band theory, which is suggested and supported by experimental data