Abstract: Methods, systems and apparatus for modeling a target system includes defining a constraint satisfaction problem (CSP) that characterizes the target system in terms of a set of variables, each having a respective input domain, and initial constraints applicable to the variables. The variables are partitioned into at least first and second sets. An abstract solution is found to the CSP, including a given assignment of the variables in the first set. A reduced domain of at least one of the variables in the second set is computed, so as to be compatible with the abstract solution. A redundant constraint on the abstract solution is determined responsively to the reduced domain. A concrete solution to the CSP is then found, using the abstract solution and the redundant constraint.

Abstract: The inventive system for estimating quantities of pollution compounds emitted in the exhaust gas of a diesel engine for a motor vehicle comprises means for regenerating a solid particle filter and an electronic control unit which manages the engine operation and is provided with one or several data memories. Said system also comprises one or several neurone networks (1) and networks (2) of input data representative for the engine operation and possibly for the vehicle motion, said data is available in the electronic control unit for managing the engine operation without an additional sensor. Said system also comprises means (4) for cumulating estimated quantities.

Abstract: A system and method for implementing an indirect controller for a plant. A plant can be provided with both a direct controller and an indirect controller with a system model or a committee of system models. When the system model has sufficient integrity to satisfy the plant requirements, i.e., when the system model has been sufficiently trained, the indirect controller with the system model is automatically enabled to replace the direct controller. When the performance falls, the direct controller can automatically assume operation of the plant, preferably maintaining operation in a control region suitable for generating additional training data for the system model. Alternatively, the system model incorporates a committee of models. Various types of sources for errors in the committee of models can be detected and used to implement strategies to improve the quality of the committee.

Abstract: A method and system is provided for analyzing a package having components. The package may include electrical or computer components. The method and system uses a computer program to receive inputted data and extract data from files. The computer program also selects the best sub-program to analyze the data to compute the parameters for packaging the components and designing the package, and displays the results of the analysis.

Abstract: A multiobjective evolutionary algorithm unit feeds a set of parameters of an individual to a search history storage device in a fitness estimating unit and to an optimization target. The optimization target outputs a set of sampled values of fitnesses on the basis of the set of parameters of the individual. The search history storage device stores the set of parameters of the individual and a set of sampled values as a search history. The fitness estimating module computes a set of estimated values of true fitnesses on the basis of the search history stored in the search history storage device for output to the multiobjective evolutionary algorithm unit. The multiobjective evolutionary algorithm unit determines a Pareto-optimal population in accordance with a genetic algorithm on the basis of a plurality of sets of estimated values.

Abstract: A method is provided to match an unknown data point with a known data point contained in a multi-dimensional data structure. The method may include receiving data from any multi-dimensional source where a query may be used to locate specific data points within that source. The method receives a number of inputs, including a Euclidean error distance and a number of reference points to use. Furthermore, the method determines optimal reference points to locate a relatively small number of data points within the data structure that possibly match the unknown point. Once possible match points are located, the method then determines the unknown point's match.

Abstract: It is difficult to select parameter values for constraint programming problem solvers which will yield good performance. Automated tuning of such problem solvers on a per problem instance basis may be used and this involves learning a function for predicting the runtime of a problem solver depending on parameter values of the problem solver and features of the problem instance being solved. However, it takes time for such prediction functions to be learnt, either during operation of a problem solver or offline, using specified examples. To address this, information about such a prediction function is shared between two or more problem solvers to improve performance. A sharing system may be used to receive prediction function information and send this to problem solvers.

Abstract: Provided is a method for determining a recovery schedule. The method includes accepting as input a recovery graph. The recovery graph presents one or more strategies for data recovery. In addition, at least one objective is provided and accepted. The recovery graph is formalized as an optimization problem for the provided objective. When formalized as an optimization problem, at least one solution technique is applied to determine at least one recovery schedule.

Abstract: A network testing environment includes a control server and a testing cluster composed of one or more load generating devices. The load generating devices output network communications in a non-deterministic manner to model real-world network users and test a network system. The load generating devices operate in accordance with probabilistic state machines distributed by the control server. The probabilistic state machines model patterns of interaction between users and the network system.

Abstract: A computer program product is described for solving the traveling salesman problem in polynomial time. The probability distribution of the space of all paths is modeled in a configurational density distribution. A Hamiltonian is constructed specifying the costs, distance, or penalty associated with different legs of paths encompassed in the configurational density distribution. Starting at a maximum temperature where free energy dominates and the penalty function plays little role, the system is iteratively adapted to reduce the temperature in steps incrementally chosen to preserve the linear characteristic of the approximation, until a lower temperature state of reduced energy is reached in which a preferred set of paths can be identified from the configurational density distribution.

Abstract: The invention relates to an evolutionary optimization method. First, an initial population of individuals is set up and an original fitness function is applied. Then the offspring individuals having a high evaluated quality value as parents are selected. In a third step, the parents are reproduced to create a plurality of offspring individuals. The quality of the offspring individuals is evaluated selectively using an original fitness function or an approximate fitness function. Finally, the method returns to the selection step until a termination condition is met. The step of evaluating the quality of the offspring individuals includes grouping all offspring individuals in clusters, selecting for each cluster one or a plurality of offspring individuals, resulting in altogether selected offspring individuals, evaluating the selected offspring individuals by the original fitness function, and evaluating the remaining offspring individuals by means of the approximate fitness function.

Abstract: A method and system for solving satisfiability problems is disclosed. In one embodiment, clauses in a satisfiability problem are organized as a chronologically ordered stack. In another embodiment, the activity of each variable in the satisfiability problem is monitored. An activity counter is maintained for each variable and is incremented each time the variable appears in a clause used in generating a conflict clause. In an embodiment, a branching variable is selected from among the variables in the top clause of the stack when the top clause is a conflict clause. In a further embodiment, one or more conflict clauses in the stack are removed when the search tree is abandoned. In a still further embodiment, the value assigned to a branching variable is selected for purposes of having a uniform distribution of positive and negative literals.

Abstract: A method and an apparatus for predicting intake manifold pressure are presented, to compensate for a large lag or a large time delay without producing an overshot or discontinuous behaviors of a predicted value. The method comprises the step of obtaining a difference of values of a variable to be predicted and a difference of values of another variable ahead of the variable to be predicted. The method further comprises the step of filtering the differences with adaptive filters. The method further comprises the step of obtaining a predicted difference of values of the variable to be predicted, through algorithm of estimation with fuzzy reasoning. The method further comprises the step of adding the predicted difference of values of the variable to be predicted, to a current value of the variable to be predicted, to obtain a predicted value of the variable to be predicted.

Abstract: A computer-implemented method and system for building a neural network is disclosed. The neural network predicts at least one target based upon predictor variables defined in a state space. First, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. In the state space, a number of points is inserted in the state space based upon the values of the predictor variables. The number of points is less than the number of observations. A statistical measure is determined that describes a relationship between the observations and the inserted points. Weights and activation functions of the neural network are determined using the statistical measure.

Abstract: An invention forcing an aggregate risk model to be consistent with standalone models is provided. A revising transformation parameterized over and an objective function minimized over, the orthogonal group are provided, least changing cross blocks of covariance matrices, preserving information in original cross block correlations, consistent with a prescribed revised sub-block.

Abstract: A method of scheduling network resources in a first domain by transforming the problem to a second domain, solving the problem and transforming back to the first domain.

Abstract: Method and apparatus for training a system model with gain constraints. A method is disclosed for training a steady-state model, the model having an input and an output and a mapping layer for mapping the input to the output through a stored representation of a system. A training data set is provided having a set of input data u(t) and target output data y(t) representative of the operation of a system. The model is trained with a predetermined training algorithm which is constrained to maintain the sensitivity of the output with respect to the input substantially within user defined constraint bounds by iteratively minimizing an objective function as a function of a data objective and a constraint objective. The data objective has a data fitting learning rate and the constraint objective has constraint learning rate that are varied as a function of the values of the data objective and the constraint objective after selective iterative steps.

Abstract: A computer implemented method for solving linear arithmetic constraints that combines a solver for difference constraints with a general linear arithmetic constraint solver. When used to solve sparse linear arithmetic constraints, the time and space complexity of the process is determined by the difference constraint component.

Abstract: Methods and apparatus for optimizing resource allocation among data analysis functions in a classification system are provided. Each of the data analysis functions is characterized as a set of operating points in accordance with at least one of resource requirements and analysis quality. An operating point for each of the data analysis functions is selected in accordance with one or more constraints. The analysis functions are applied at selected operating points to optimize resource allocation among the data analysis functions in the classification system.

Abstract: Techniques are presented for identifying blockable subsets. Blockable subsets can increase the efficiency by which solutions to a constraint set representation (CSR) can be found. Nodes of a blockable subset can be marked as “blocked” and learning or implication procedures, used as part of a CSR solving process, can be designed to skip nodes marked as blocked. The identification of a particular blockable subset is typically associated with certain conditions being true. If and when the conditions no longer hold, the nodes of the blockable subset need to be unblocked. One type of blockable subset can be identified during the operation of an implication engine (IE) by a technique called justified node blocking (JNB). Another type of blockable subset can be identified by a technique called pivot node learning (PNL). PNL can be applied in-between application of an IE and application of case-based learning.

Abstract: Trade-off/semantic networks are described, including identifying a plurality of elemental attributes, relating each of the plurality of elemental attributes to each other to configure a trade-off relationship between each of the plurality of elemental attributes, and defining an agility extension configured to modify a value for each of the plurality of elemental attributes, where the value affects the trade-off relationship between each of the plurality of elemental attributes.

Abstract: One embodiment of the present invention provides a system that uses interval techniques to solve a multi-objective optimization problem. During operation, the system receives a representation of multiple objective functions (ƒ1, . . . , ƒn) at a computer system, wherein (ƒ1, . . . , ƒn) are scalar functions of a vector x=(x1, . . . , xn). The system also receives a representation of a domain of interest for the multiple objective functions. Next, the system performs an interval optimization process to compute guaranteed bounds on a Pareto front for the objective functions (ƒ1, . . . , fn), wherein for each point on the Pareto front, an improvement in one objective function cannot be made without adversely affecting at least one other objective function. While performing the interval optimization process, the system applies a direct-comparison technique between subdomains of the domain of interest to eliminate subdomains that are certainly dominated by other subdomains.

Abstract: A system and method for generating a neural network ensemble. Conventional algorithms are used to train a number of neural networks having error diversity, for example by having a different number of hidden nodes in each network. A genetic algorithm having a multi-objective fitness function is used to select one or more ensembles. The fitness function includes a negative error correlation objective to insure diversity among the ensemble members. A genetic algorithm may be used to select weighting factors for the multi-objective function. In one application, a trained model may be used to produce synthetic open hole logs in response to inputs of cased hole log data.

Abstract: In one embodiment, the present invention is a method and apparatus for automatic system parameter configuration for performance improvement. One embodiment of the inventive method involves formulating a black box optimization problem, and solving the optimization problem using an enhanced smart hill climbing method. The smart hill climbing method includes both a global and a more precise local search to identify an optimal solution.

Abstract: An intelligent spatial reasoning method receives a plurality of object sets. A spatial mapping feature learning method uses the plurality of object sets to create at least one salient spatial mapping feature output. It performs spatial reasoning rule learning using the at least one spatial mapping feature to create at least one spatial reasoning rule output. The spatial mapping feature learning method performs a spatial mapping feature set generation step followed by a feature learning step. The spatial mapping feature set is generated by repeated application of spatial correlation between two object sets. The feature learning method consists of a feature selection step and a feature transformation step and the spatial reasoning rule learning method uses the supervised learning method. The spatial reasoning approach of this invention automatically characterizes spatial relations of multiple sets of objects by comprehensive collections of spatial mapping features.

Abstract: A Soft Computing (SC) optimizer for designing a Knowledge Base (KB) to be used in a control system for controlling a motorcycle is described. In one embodiment, a simulation model of the motorcycle and rider control is used. In one embodiment, the simulation model includes a feedforward rider model. The SC optimizer includes a fuzzy inference engine based on a Fuzzy Neural Network (FNN). The SC Optimizer provides Fuzzy Inference System (FIS) structure selection, FIS structure optimization method selection, and teaching signal selection and generation. The user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference; and the preliminary type of membership functions. A Genetic Algorithm (GA) is used to optimize linguistic variable parameters and the input-output training patterns. A GA is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. The GA produces a near-optimal FNN.

Abstract: In the implementation of a production process, event data corresponding to a parent event that has triggered an executable business process that has failed is logged. The event data is displayed to a user. Responsive to input from the user, a child event based on the event data is generated, and the child event is submitted to a process engine to initiate re-execution of the executable business process in accordance with the child event.

Abstract: The present invention involves a Soft Computing (SC) optimizer for designing a Knowledge Base (KB) to be used in a control system for controlling a plant such as, for example, an internal combustion engine or an automobile suspension system. The SC optimizer includes a fuzzy inference engine based on a Fuzzy Neural Network (FNN). The SC Optimizer provides Fuzzy Inference System (FIS) structure selection, FIS structure optimization method selection, and teaching signal selection and generation. The user selects a fuzzy model, including one or more of: the number of input and/or output variables; the type of fuzzy inference model (e.g., Mamdani, Sugeno, Tsukamoto, etc.); and the preliminary type of membership functions. A Genetic Algorithm (GA) is used to optimize linguistic variable parameters and the input-output training patterns. A GA is also used to optimize the rule base, using the fuzzy model, optimal linguistic variable parameters, and a teaching signal. The GA produces a near-optimal FNN.

Abstract: A method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. The method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite Radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output.

Abstract: In a method for determining a minimum value of an optimization function under constraints given by equations, a set of points which satisfy the constraints is regarded as a Riemannian manifold within a finite-dimensional real-vector space, the Riemannian manifold is approached from an initial position within the real-vector space. An exponential map regarding a geodesic line equation with respect to a tangent vector on the Riemannian manifold ends at a finite order, an approximate geodesic line is generated as a one-dimensional orbit. An approximate parallel-translation is performed on the tangent vector on the Riemannian manifold and on the orbit generated in the orbit generating step by finite-order approximation of the exponential map regarding the parallel translation of the tangent vector.

Abstract: The present invention provides a method of optimizing a painting process for applying a paint layer on an article. The method comprises defining a functional relationship paint processing parameters and a paint layer property (i.e., the average paint layer thickness) using a neural network. This functional relationship is then used in a paint optimization function that measures a combination of quality control parameters and efficiency parameters. Finally, the paint optimization function is optimized by adjusting the paint processing parameters utilizing the functional relationship formed by the neural network. The invention also provides a system that implements the methods of the invention.

Abstract: A method for solving a constraint satisfaction problem (CSP) includes choosing a first state corresponding to a first set of values of a set of variables, and selecting a hop distance within a state space of the variables responsively to a random distance selection criterion. A second state corresponding to a second set of the values of the variables is selected, such that the second state is separated from the first state by the hop distance. Constraint costs of the first and second states are compared. If the cost of the second state is lower than the cost of the first state, the first state is redefined to correspond to the second set of the values of the variables. These steps are repeated until a solution of the CSP is found.

Abstract: A computer-implemented method and system for building a neural network is disclosed. The neural network predicts at least one target based upon predictor variables defined in a state space. First, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. In the state space, a number of points is inserted in the state space based upon the values of the predictor variables. The number of points is less than the number of observations. A statistical measure is determined that describes a relationship between the observations and the inserted points. Weights and activation functions of the neural network are determined using the statistical measure.

Abstract: In one embodiment, a system and method for solving linear programs includes a perceptron algorithm configured to move toward a solution to the linear program. A transform algorithm is configured to stretch portions of a vector space within which the linear program is defined. A decision module decides between continued application of the perceptron algorithm and application of the transform algorithm based on a rate at which the approximate solutions are approaching a satisfactory solution.

Abstract: A method for changing the CPU frequency under control of a neural network. The neural network has m basis functions and n basis points that are connected together. Using the learning capability of the neural network to deduce basis weights based on dummy environmental parameters and a dummy output vector. In an application procedure, environmental parameters are input to the basis points and basis vectors are calculated based on the basis functions. Integrating the multiplication of each basis vector and its corresponding basis weight, an output vector can be generated to determine a control signal so that the CPU can be controlled to raise or lower its operating frequency. In addition, if the user has to change the parameters due to behavior, a fast learning function of a radial neural network can be used for complying with each user's behavior.

Abstract: Non-linear model with disturbance rejection. A method for training a non linear model for predicting an output parameter of a system is disclosed that operates in an environment having associated therewith slow varying and unmeasurable disturbances. An input layer is provided having a plurality of inputs and an output layer is provided having at least one output for providing the output parameter. A data set of historical data taken over a time line at periodic intervals is generated for use in training the model. The model is operable to map the input layer through a stored representation to the output layer. Training of the model involves training the stored representation on the historical data set to provide rejection of the disturbances in the stored representation.

Abstract: A method, apparatus and article of manufacture for problem identification and resolution using intelligent agents. In at least one embodiment, an agent is a software element configured to detect a situation (e.g., problem or problems) and take steps to preserve a context in which the situation occurs. The agent may also be configured to identify one or more courses of action (e.g., solutions) to be taken in response to the situation. In one embodiment, a user trains an agent to take a particular action upon detecting a particular problem. The training may be initiated after accessing a log containing details about the problem context and recommended courses of action.

Abstract: The methods, systems and devices of the present invention comprise use of Support Vector Machines and RFE (Recursive Feature Elimination) for the identification of patterns that are useful for medical diagnosis, prognosis and treatment. SVM-RFE can be used with varied data sets.

Abstract: One embodiment of the present invention provides a system that solves a global optimization problem specified by a function ƒ and a set of equality constraints q1(x)=0 (i=1, . . . , r), wherein ƒ is a scalar function of a vector x=(x1, x2, x3, . . . xn). During operation, the system receives a representation of the function ƒ and the set of equality constraints and stores the representation in a memory within a computer system. Next, the system and performs an interval global optimization process to compute guaranteed bounds on a globally minimum value of the function ƒ(x) subject to the set of equality constraints. Performing this interval global optimization process involves, applying term consistency to the set of equality constraints over a subbox X, and excluding portions of the subbox X that violate the set of equality constraints.

Abstract: A method for action selection based upon an objective of an outcome relative to a subject. In one embodiment, a training set is obtained that contains attributes of a subject. In the present embodiment, a best behavioral model for predicting an outcome when a subject has an action applied is calculated. The training set is mapped to the best behavioral model. The mapping provides a base from which a random sub-sample is acquired. In the present embodiment, a random sub-sample of the training set and the best behavioral model is then selected. This random sub-sample reduces the computational requirements when determining an optimized strategy. The optimized strategy provides an optimal action relative to the subject for the objective of the outcome. In one embodiment, the subject is a customer of a business entity, enabled to interact with the customer, and an action is a promotion offered by the business entity.

Abstract: A method for feedback control of cooperative problem solving for real-time applications in complex systems utilizes solvers parameterized by control variables. The method includes initializing the time setting and selecting at least one solver parameter value. The solver is operated with the selected solver parameter value or values for a specified interim and the operational conditions are reviewed. A solution is transmitted to the system if a solution quality condition is satisfied. The solver continues to operate if the solution quality condition is not satisfied and the performance differential is not greater than a specified threshold. If the solution quality condition is unsatisfied, but the performance differential exceeds the threshold, at least one alternate solver parameter value is selected and the solver is operated with the new solver parameter value for a specified interim. The solver continues to operate until the solution quality condition is satisfied.

Abstract: A cooperative solving method for controlling a plurality of constraint problem solvers identifies complexity criteria, which provide direction for selecting and for transitioning between constraint problem solvers. The method includes randomly selecting a test point and determining whether the test point satisfies a first complexity criterion. A first constraint problem solver is selected, and an alternate test point is identified by the first solver if the complexity criterion has not been satisfied. If the alternate test point is a problem solution, it is transmitted to the system. If the alternate test point is not a problem solution or if the original randomly-selected test point satisfies the complexity criterion, a second constraint solver selects a new test point. If the new test point is a problem solution, it is transmitted to the system; if the new test point is not a solution, the cooperative solver is restarted.

Abstract: Method and apparatus for training a system model with gain constraints. A method is disclosed for training a steady-state model, the model having an input and an output and a mapping layer for mapping the input to the output through a stored representation of a system. A training data set is provided having a set of input data u(t) and target output data y(t) representative of the operation of a system. The model is trained with a predetermined training algorithm which is constrained to maintain the sensitivity of the output with respect to the input substantially within user defined constraint bounds by iteratively minimizing an objective function as a function of a data objective and a constraint objective. The data objective has a data fitting learning rate and the constraint objective has constraint learning rate that are varied as a function of the values of the data objective and the constraint objective after selective iterative steps.

Abstract: Dynamical methods for obtaining the global optimal solution of general optimization problems having closed form or black box objective functions, including the steps of first finding, in a deterministic manner, one local optimal solution starting from an initial point, and then finding another local optimal solution starting from the previously found one until all the local optimal solutions starting from any initial point are found, and then finding from said points the global optimal solution.

Abstract: A method for practical QoS routing, which provides a solution to the delay constrained least cost routing problem, is presented. The method uses the concept of aggregated costs and finds the optimal multiplier based on Lagrange relaxation. The method is polynomial in running time, and produces a theoretical lower bound (i.e. optimal solution), along with the result. The differences between the lower bound and result are small, indicating the quality of the result. Additionally, by further relaxing the desire for an optimal solution, an option is provided to control the trade-off between running time of the algorithm and quality of the result.

Abstract: Computer-implemented methods, systems and apparatus, including computer program apparatus, provide techniques for designing a set of experiments to be performed with a set of resources. A plurality of experimental configurations are generated based on a set of parameters describing factors to be varied in the experiments and a set of constraints representing limitations on operations that can be performed with the set of resources. A set of experiments is defined based on a selected configuration. The constraints can be represented as patterns defining an application of a parameter to a set of one or more points of an experimental lattice.

Abstract: A method for developing and using real time applications for a dynamic system having a sensing subsystem, actuation subsystem, a control subsystem, and an application subsystem utilizes stochastic compute time algorithms. After optimization functions, desired state and constraints are received and detector data has been provided from a sensor subsystem, a statistical optimization error description is generated. From this statistical optimization error description a strategy is developed, including the optimization errors, within the control subsystem. An execution module within the control subsystem then sends an execution strategy to various actuators within the actuation subsystem.

Abstract: An automated system adaptation technique for computer systems, networks and subsystems generally and, more particularly, for data storage systems. The invention programmatically designs, configures and manages a system, such as a data storage system. This is accomplished by performing a sequence of steps in an iterative loop, including analyzing the operation of the system under a workload, generating a new design based on the analysis and migrating the existing system to the new design. By systematically exploring a large design space and developing designs based on analyses of the workload, the invention generates designs that are improved in comparison to conventional design techniques. By programmatically repeating these tasks, the invention causes the system to converge to one that supports the workload without being over-provisioned.

Abstract: A method of effectively designing an interactive system is disclosed. The method may preferably include the steps of (1) determining the overall scope of the desired interactive system, including available sources of input and output; (2) creating a list of desired features for the system; (3) creating a rough outline of perceptions, decisions, and actions that the system must be capable of if it is to possess the desired features; (4) designing perception features that implement the identified perception specifications and decision features that implement the decision-making process required by the identified decision specifications; (5) creating any necessary supporting components (e.g., buffers for storing data collected by perception features) to implement the perception and decision features; and (6) creating a behavior feature hierarchy that includes the action features required to implement action features of the interactive system.

Abstract: Complex process control and maintenance are performed utilizing a nonlinear regression analysis to determine optimal maintenance activities and process adjustments based on an urgency metric.