Abstract: Partial differential equations used to simulate physical systems can be solved, in one embodiment, by a solver that has been trained with a set of generative neural networks that operated at different resolutions in a solution space of a domain that defines the physical space of the physical system. The solver can operate in a latent vector space which encodes solutions to the PDE in latent vectors in the latent vector space. The variables of the PDE can be partially decoupled in the latent vector space while the solver operates. The domain can be divided into subdomains that are classified based on their positions in the domain.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using neural networks having contiguous sparsity patterns. One of the methods includes storing a first parameter matrix of a neural network having a contiguous sparsity pattern in storage associated with a computing device. The computing device performs an inference pass of the neural network to generate an output vector, including reading, from the storage associated with the computing device, one or more activation values from the input vector, reading, from the storage associated with the computing device, a block of non-zero parameter values, and multiplying each of the one or more activation values by one or more of the block of non-zero parameter values.

Abstract: The present disclosure advantageously provides a heterogenous system, and a method for generating an artificial neural network (ANN) for a heterogenous system. The heterogenous system includes a plurality of processing units coupled to a memory configured to store an input volume. The plurality of processing units includes first and second processing units. The first processing unit includes a first processor and is configured to execute a first ANN, and the second processing unit includes a second processor and is configured to execute a second ANN. The first and second ANNs respectively include an input layer, at least one processor-optimized hidden layer and an output layer. The second ANN hidden layers are different than the first ANN hidden layers.

Abstract: A large amount of training data is typically required to perform deep network leaning, making it difficult to achieve using a few pieces of data. In order to solve this problem, the neural network device according to the present invention is provided with: a feature extraction unit which extracts features from training data using a learning neural network; an adversarial feature generation unit which generates an adversarial feature from the extracted features using the learning neural network; a pattern recognition unit which calculates a neural network recognition result using the training data and the adversarial feature; and a network learning unit which performs neural network learning so that the recognition result approaches a desired output.

Abstract: Techniques for training a machine learning model are described herein. For example, the techniques may include implementing a cross batch normalization layer that generates a cross batch normalization layer output based on a first layer output during training of the neural network. The training may be based on a local batch of training examples of a global batch including the local batch and at least one remote batch of training examples. The cross batch normalization layer output may include normalized components of the first layer output determined based on global normalization statistics for the global batch. Such techniques may be used to train a neural network over distributed machines by synchronizing batches between such machines.

Abstract: Methods, systems, and computer-readable storage media for a machine learning (ML) model and framework for training of the ML model to enable the ML model to correctly match entities even in instances where new entities are added after the ML model has been trained. More particularly, implementations of the present disclosure are directed to a ML model provided as a neural network that is trained to provide a scalar confidence score that indicates whether two entities in a pair of entities are considered a match, even if an entity in the set of entities was not accounted for in training of the ML model.

Abstract: A computer-implemented method for power grid anomaly detection using a convolutional neural network (CNN) trained to detect anomalies in electricity demand data and electricity supply data includes receiving (i) electricity demand data comprising time series measurements of consumption of electricity by a plurality of consumers, and (ii) electricity supply data comprising time series measurements of availability of electricity by one or more producers. An input matrix is generated that comprises the electricity demand data and the electricity supply data. The CNN is applied to the input matrix to yield a probability of anomaly in the electricity demand data and the electricity supply data. If the probability of anomaly is above a threshold value, an alert message is generated for one or more system operators.

Abstract: Methods for multi-component topology optimization for composite structures are disclosed. In one embodiment, a method of designing a structure by computer-implemented topology optimization includes establishing a plurality of design points within a design domain and establishing at least a first orientation field and a second orientation field. The method further includes assigning values for the one or more membership fields, the one or more density fields, the first orientation field and the second orientation field, and projecting the values onto a simulation model. The method includes achieving convergence of an objective function for a design variable by iteratively executing a topology optimization of the simulation model using the values. Each design point of the plurality of design points is a member of no component or a member of one of the first component and the second component.

Abstract: Improved convolutional neural network-based machine learning models are disclosed herein. A convolutional neural network is configured to decompose feature maps generated based on a data item to be classified. The feature maps are decomposed into a first and second subsets. The first subset is representative of high frequency components of the data item, and the second subset is representative of low frequency components of the data item. The second subset is upsampled and is combined with the first subset. The combined feature maps are convolved with a filter to extract a set of features associated with the data item. The first subset is also downsampled and combined with the second subset. The combined feature maps are convolved with a filter to extract another set of features. The data item is classified based on the sets of features extracted based on the convolution operations.

Abstract: Systems and methods for mining hybrid automata from input-output traces of cyber-physical systems are disclosed herein.

Abstract: Embodiments include methods, and computer system, and computer program products for assessing and remediating online servers with minimal impact. Aspects include: duplicating, in real-time at time T0, first instance of computer resources of first server into second instance of computer resources of second server, the first instance of computer resources having first instance of operating systems, first instance of applications and first instance of data and the second instance of computer resources having second instance of operating systems, second instance of applications and second instance of data, running assessment and remediation on the second instance of operating systems and applications of the second server, merging the second instance of data of the second server with the first instance of data of the first server, and swapping the identities of the first instance of computer resources of the first server and the second instance of computer resources of the second server.

Abstract: Methods and apparatus are provided for using a breakpoint determination unit to examine an artificial nervous system. One example method generally includes operating at least a portion of the artificial nervous system; using the breakpoint determination unit to detect that a condition exists based at least in part on monitoring one or more components in the artificial nervous system; and at least one of suspending, examining, modifying, or flagging the operation of the at least the portion of the artificial nervous system, based at least in part on the detection.

Abstract: A discrete element method for modelling granular or particulate material, the method including a multiple grid search method wherein the multiple grid search method is a hierarchical grid search method, and wherein entities, such as particles and boundary elements, are allocated to cells of respective grids based on size. The search method further includes: (a) performing a search of cells in a first of the grid levels to determine pairs of entities which satisfy predetermined criteria to be included in a neighbor list for which both entities belong to the first grid level; (b) mapping each nonempty cell in the first grid level to each of the other grid levels, determining neighboring cells in each of the other grid levels and determining all pairs of entities belonging to pair of levels that satisfy the predetermined criteria for inclusion in the neighbor list; and (c) repeating (a) and (b) for all grid levels.

Abstract: Exemplary methods, apparatuses, and systems receive data as input to be parsed. The data is parsed using a plurality of pattern matching rules, the plurality of pattern matching rules organized according to a hierarchy including a parent rule and one or more child rules of the parent rule. Parsing includes applying the parent rule to the unstructured data, determining the parent rule is unable to find a pattern match in the unstructured data, and bypassing the application of each child rule to the unstructured data in response to the determination that the parent rule is unable to find a pattern match.

Abstract: Distributed mobile device management including a plurality of management agents is disclosed. Management-related information may be retrieved from a storage location accessible to a plurality of management agents. The management-related information may have been provided to the storage location from a management agent associated with a managed application. And at least one operation may be performed based at least in part on the management-related information.

Abstract: A multivariate digital camera device and method for generating pictures of datasets comprised of points in hyperspace is provided. The invention may be embodied in an input device, a computer processor, and an output device. The input device may be a keyboard, a laboratory instrument such as a mass spectrometer, a reader of computer readable medium, or a network interface device. The output device may be a monitor used in conjunction with either a 2D or 3D printer or both. The computer processor receives data from the input device and performs a series of steps to create a 2D or 3D image of the hyperspace object. The resulting image is then produced in a non-transitory medium by at least one of the output devices. The processor steps include the use of a maximum distance method in which distances and angles information from the points in the hyperspace dataset are preserved to produce a more picture-like quality.

Abstract: A prediction method that estimates the real-time position of a mobile device based on previously observed data is provided. The present invention can be used in real-time navigation, including providing real-time alerts of an upcoming destination and notifications of emergency events in close geographic proximity. The prediction method utilizes neural networks and/or functions generated using genetic algorithms in estimating the mobile device's real-time position. The prediction method provides reliable Location-Based Services (LBS) in events where traditional positioning technologies become unreliable. It is also seamless, as the user remains unaware of any interruption in accessing the positioning technology.

Abstract: A catheterization device that may be designed by use of an adaptive genetic algorithm computational fluid dynamics approach, as well as other Global Optimization methods that may include simulated annealing, multistart and interval methods, continuous branch and bound methods, evolutionary algorithms, and tabu search and scatter search methods, as well as other available Global Optimization methods that is able to maximize/optimize the dwell time of an infused agent in the vicinity of a vascular lesion. The device may have an internal by-pass channel that allows the blood upstream of the lesion to continue its pulsatile flow through the vessel in the part of it occluded by the lesion, while simultaneously allowing the disbursement and maximal dwell time of an antithrombolytic or other diagnostic or therapeutic agent needed to treat the lesion.

Abstract: A neuro-fuzzy controller is provided. The neuro-fuzzy controller includes a predictor that receives inputs and makes prediction inputs. The prediction inputs are passed to a fuzzy cluster module that includes a neural network fuzzifing said prediction inputs and passing the result to an inference engine. The output of the inference engine is defuzzified and provided as an output of the controller. The fuzzifier and defuzzifier preferably represent a neural network employing a trigonometrical series. The inference engine preferably employs rules that are determined using genetic programming.

Abstract: Disclosed herein are systems, methods, and computer readable-media for contextual adaptive advertising. The method for contextual adaptive advertising comprises tracking user behavior across multiple modalities or devices, storing one or more representations of user behavior in a log as descriptors, normalizing the descriptors, merging the normalized descriptors into a unified click or interactive stream, and generating a behavioral model by analyzing the click or interactive stream.

Abstract: A prediction method that estimates the real-time position of a mobile device based on previously observed data is provided. The present invention can be used in real-time navigation, including providing real-time alerts of an upcoming destination and notifications of emergency events in close geographic proximity. The prediction method utilizes neural networks and/or functions generated using genetic algorithms in estimating the mobile device's real-time position. The prediction method provides reliable Location-Based Services (LBS) in events where traditional positioning technologies become unreliable. It is also seamless, as the user remains unaware of any interruption in accessing the positioning technology.

Abstract: Certain embodiments of the present invention provide a system, method and computer instructions for aiding analysis of an image used in a medical examination. An image analysis system used in a medical examination includes an input module configured to input an image a search module configured to locate information regarding an image that is similar to the input image and an output module configured to output a link to the located information, wherein the located information is displayed when the link is used. The image analysis system used in the medical examination further includes a communication module, wherein the communication module is configured to output the input image and the located information, and wherein the communication module is configured to receive a responsive communication.

Abstract: Methods, systems, and computer readable media are provided for fast updating of oil and gas field production optimization using physical and proxy simulators. A base model of a reservoir, well, or a pipeline network is established in one or more physical simulators. A decision management system is used to define uncertain parameters for matching with observed data. A proxy model is used to fit the uncertain parameters to outputs of the physical simulators, determine sensitivities of the uncertain parameters, and compute correlations between the uncertain parameters and output data from the physical simulators. Parameters for which the sensitivities are below a threshold are eliminated. The decision management system validates parameters which are output from the proxy model in the simulators. The validated parameters are used to make production decisions.

Abstract: A neuromorphic computing device utilizing electronics to perform the function of neurons and synaptic connections. The invention provides variable resistance circuits to represent interconnection strength between neurons and a positive and negative output circuit to represent excitatory and inhibitory responses, respectively. The invention provides advantages over software-based neuromorphic computing methods.

Abstract: A process simulator is provided for simulating the behavior of multi-dimensional non-linear multivariable processes. A multi-dimensional non-linear multivariable model of a process can be generated, such as by using smaller building blocks. One or more inputs are provided to the model, a behavior of the process is simulated in real-time using the model, and one or more outputs of the model are provided. The model could represent a two-dimensional non-linear multivariable model, and the one or more inputs to the model and/or the one or more outputs of the model could be array-based. The process simulator could be formed from multiple components, such as a regulatory loop simulator object, a process model object, a disturbance generator object, and a scanner simulator object. The arrangement of the objects can be flexible and configurable, such as by designing the objects in an object-oriented manner utilizing a sink/source architecture.

Abstract: The method for generating an integrated guidance law for aerodynamic missiles uses a strength Pareto evolutionary algorithm (SPEA)-based approach for generating an integrated fuzzy guidance law, which includes three separate fuzzy controllers. Each of these fuzzy controllers is activated in a unique region of missile interception. The distribution of membership functions and the associated rules are obtained by solving a nonlinear constrained multi-objective optimization problem in which final time, energy consumption, and miss distance are treated as competing objectives. A Tabu search is utilized to build a library of initial feasible solutions for the multi-objective optimization algorithm. Additionally, a hierarchical clustering technique is utilized to provide the decision maker with a representative and manageable Pareto-optimal set without destroying the characteristics of the trade-off front. A fuzzy-based system is employed to extract the best compromise solution over the trade-off curve.

Abstract: The control optimization method for helicopters carrying suspended loads during hover flight utilizes a controller based on time-delayed feedback of the load swing angles. The controller outputs include additional displacements, which are added to the helicopter trajectory in the longitudinal and lateral directions. This simple implementation requires only a small modification to the software of the helicopter position controller. Moreover, the implementation of this controller does not need rates of the swing angles. The parameters of the controllers are optimized using the method of particle swarms by minimizing an index that is a function of the history of the load swing. Simulation results show the effectiveness of the controller in suppressing the swing of the slung load while stabilizing the helicopter.

Abstract: The fuzzy logic-based control method for helicopters carrying suspended loads utilizes a controller based on fuzzy logic membership distributions of sets of load swing angles. The anti-swing controller is fuzzy-based and has controller outputs that include additional displacements added to the helicopter trajectory in the longitudinal and lateral directions. This simple implementation requires only a small modification to the software of the helicopter position controller. The membership functions govern control parameters that are optimized using a particle swarm algorithm. The rules of the anti-swing controller are derived to mimic the performance of a time-delayed feedback controller. A tracking controller stabilizes the helicopter and tracks the trajectory generated by the anti-swing controller.

Abstract: Certain embodiments of the present invention provide a system, method and computer instructions for aiding analysis of an image used in a medical examination. For example, in an embodiment, an image analysis system used in a medical examination comprises: an input module configured to input an image; a search module configured to locate information regarding an image that is similar to the input image; and an output module configured to output a link to the located information, wherein the located information is displayed when the link is used. For example, in an embodiment, an image analysis system used in a medical examination further comprises a communication module, wherein the communication module is configured to output the input image and the located information, and wherein the communication module is configured to receive a responsive communication.

Abstract: Outlier detection methods and apparatus have light computational resources requirement, especially on the storage requirement, and yet achieve a state-of-the-art predictive performance. The outlier detection problem is first reduced to that of a classification learning problem, and then selective sampling based on uncertainty of prediction is applied to further reduce the amount of data required for data analysis, resulting in enhanced predictive performance. The reduction to classification essentially consists in using the unlabeled normal data as positive examples, and randomly generated synthesized examples as negative examples. Application of selective sampling makes use of an underlying, arbitrary classification learning algorithm, the data labeled by the above procedure, and proceeds iteratively.

Abstract: A packet-network analyzer system for characterizing network conditions of a packet-network-under-test is provided. In this regard, one such system can be broadly summarized by a representative analyzer system that incorporates a data collection element to receive the raw digital data from a host analyzer, a data selection element to receive the raw digital data, a data processing element to process the selected data set to generate a normalized data set, a neural processing module to process the normalized data set to generate a set of rules and relationships, and a data mining module that uses the rules and relationships to generate a mined data set from the selected data set, the mined data set being used to characterize a packet-network-under test.

Abstract: A global software development model instrument is described. The instrument utilizes a global system dynamics model, as well as one or more site-specific discrete event simulation and system dynamics models to model interactions within and between software development sites. Parameters, equations, and interactions between the model components are editable to allow for the simulation and comparison of various software development options and to provide for global software development research. Additional product development situations can be modeled as well, including hardware and systems engineering.

Abstract: In an active system, an actor is able to effect action in a subject system. The actor and the subject system exist in an environment which can impact the subject system. Neither the actor nor the subject system has any control over the environment. The actor includes a model and a processor. The processor is guided by the model. The processor is arranged to effect action in the subject system. The subject system is known by the model. This allows the actor to be guided in its action on the subject system by the model of the subject system. Events can occur in the subject system either through the actions of the actor, as guided by the model, or through actions of other actors, or through a change in state of the subject system itself (e.g. the progression of a chemical reaction) or its environment (e.g. the passage of time). The actor keeps the model updated with its own actions.

Abstract: An outlier detector that exploits the existing risk structure of the decision problem in order to discover risk assignments that are globally inconsistent is described. The technique works on a set of candidates for which risk categories have already been assigned. In the case of insurance underwriting, the invention pertains to the premium class assigned to an application. For this set of labeled candidates, the system finds all such pairs of applications belonging to different risk categories, which violate the principle of dominance. The invention matches the risk ordering of the applications with the ordering imposed by dominance and uses any mismatch during the process to identify applications that were potentially assigned incorrect risk categories.

Abstract: The present invention provides systems and methods for dynamic detection and prevention of electronic fraud and network intrusion using an integrated set of intelligent technologies. The intelligent technologies include neural networks, multi-agents, data mining, case-based reasoning, rule-based reasoning, fuzzy logic, constraint programming, and genetic algorithms. The systems and methods of the present invention involve a fraud detection and prevention model that successfully detects and prevents electronic fraud and network intrusion in real-time. The model is not sensitive to known or unknown different types of fraud or network intrusion attacks, and can be used to detect and prevent fraud and network intrusion across multiple networks and industries.

Abstract: A work model (or an image) is displayed on an image plane of a robot simulator (201), and a measuring portion and a measuring method are designated (202, 203) and a work shape and a work loading state are designated (204), and then it is judged whether or not the measuring portion and the measuring method are good (205). When the measuring portion and the measuring method are good, a program is generated and the processing is completed (207, 208). When the measuring portion and the measuring method are not good, an alarm is given (206), and the continuation (207) or the repetition (201) of the processing is directed. At the time of analyzing the program, the loading (101), the analysis and display of the measuring portion and the measuring method (102, 103) and the work information (104) are designated, and then it is judged whether or not the measuring portion and the measuring method, which have been analyzed, are good (105).

Abstract: Systems, methods, and computer program products implementing techniques for training classifiers. The techniques include receiving a training set that includes positive samples and negative samples, receiving a restricted set of linear operators, and using a boosting process to train a classifier to discriminate between the positive and negative samples. The boosting process is an iterative process. The iterations include a first iteration where a classifier is trained by (1) testing some, but not all linear operators in the restricted set against a weighted version of the training set, (2) selecting for use by the classifier the linear operator with the lowest error rate, and (3) generating a re-weighted version of the training set. The iterations also include subsequent iterations during which another classifier is trained by repeating steps (1), (2), and (3), but using in step (1) the re-weighted version of the training set generated during a previous iteration.

Abstract: We develop a system consisting of a neural architecture resulting in classifying regions corresponding to users' keystroke patterns. We extend the adaptation properties to classification phase resulting in learning of changes over time. Classification results on login attempts of 43 users (216 valid, 657 impersonation samples) show considerable improvements over existing methods.

Abstract: A method, computer program product, and system for variant string matching. A computer implemented method for variant string matching may comprise comparing with a computing device two unidentical strings in a training variant string pair. The two unidentical strings may represent the same item from training data, which may be stored in a memory. The two unidentical strings may be compared to determine if they include an identical substring pair, and a first unidentical substring pair. The computer implemented method may also determine if the first unidentical substring pair includes a first unidentical substring and a second unidentical substring. The computer implemented method may further determine if the first unidentical substring pair is in the training data. The first unidentical substring pair may be entered into the training data as a first variant string pair if it is not in the training data.

Abstract: A risk classification technique that exploits the existing risk structure of the decision problem in order to produce risk categorizations for new candidates is described. The technique makes use of a set of candidates for which risk categories have already been assigned (in the case of insurance underwriting, for example, this would pertain to the premium class assigned to an application). Using this set of labeled candidates, the technique produces two subsets for each risk category: the Pareto-best subset and the Pareto-worst subset by using Dominance. These two subsets can be seen as representing the least risky and the most risky candidates within a given risk category. If there are a sufficient number of candidates in these two subsets, then the candidates in these two subsets can be seen as samples from the two hypothetical risk surfaces in the feature space that bound the risk category from above and below respectively.

Abstract: Methods and systems for obtaining the performance characteristics of a computing product are described. Obtaining a computing product's attributes, capabilities, and features includes assessing the computing product to determine the product's attributes, capabilities, and features. Once the assessment is completed, the assessment data is recorded and stored for future applications. The assessments can be performed by the operating system through an assessment tool. Assessments can be performed on various computing products including personal computers, computer components, clusters of computers, and servers.

Abstract: Disclosed herein is a programming tool stored on a computer-readable medium and adapted for implementation by a computer for designing an artificial neural network. The programming tool includes a network configuration module to provide a first display interface to support configuration of the artificial neural network, and a pattern data module to provide a second display interface to support establishment and modification of first and second pattern data sets for training and testing the artificial neural network, respectively.

Abstract: A method of operating a plurality of electrical and/or electronic devices connected with a data processing apparatus is disclosed as including the steps of (a) detecting occurrence of events of the devices (b) recording data relating to some of said detected events in a database; (c) ordering the recorded events into sequences of events chronologically and/or geographically; (d) comparing a detected event with the sequences of events for finding a matched sequence of events; and (e) performing the remaining events in the matched sequence.

Abstract: Systems, methods, and computer program products implementing techniques for training classifiers. The techniques include receiving a training set that includes positive images and negative images, receiving a restricted set of linear operators, and using a boosting process to train a classifier to discriminate between the positive and negative images. The boosting process is an iterative process. The iterations include a first iteration where a classifier is trained by (1) testing some, but not all linear operators in the restricted set against a weighted version of the training set, (2) selecting for use by the classifier the linear operator with the lowest error rate, and (3) generating a re-weighted version of the training set. The iterations also include subsequent iterations during which another classifier is trained by repeating steps (1), (2), and (3), but using in step (1) the re-weighted version of the training set generated during a previous iteration.

Abstract: Methods and systems are disclosed herein in which a physical neural network can be configured utilizing nanotechnology. Such a physical neural network can comprise a plurality of molecular conductors (e.g., nanoconductors) which form neural connections between pre-synaptic and post-synaptic components of the physical neural network. Additionally, a learning mechanism can be applied for implementing Hebbian learning via the physical neural network. Such a learning mechanism can utilize a voltage gradient or voltage gradient dependencies to implement Hebbian and/or anti-Hebbian plasticity within the physical neural network. The learning mechanism can also utilize pre-synaptic and post-synaptic frequencies to provide Hebbian and/or anti-Hebbian learning within the physical neural network.

Abstract: A modulation recognition method and device for digitally modulated signals with multi-level magnitudes are provided. The modulation recognition method includes selecting plural quantization sizes used to construct plural statistic histograms related to the magnitude of a sequence of data, setting up an off-line processing to extract plural useful feature patterns for each modulation type of interest, receiving a sequence of samples of a modulated object signal and constructing plural statistic histograms related to the magnitude of these samples, and adopting a hierarchical classification method for modulation recognition. It can be applied to the adaptive-modulation communication system, software defined radio, digital broadcasting systems and military communication systems. It can also be integrated with modulation recognition techniques for other types of modulated signals to function in a universal demodulator.

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: Software (100, 600, 1000, 1100) for automatically designing and optimizing signal processing networks (e.g., 200, 700, 800, 900) is provided. The software use genetic programming e.g., gene expression programming in combination with numerical optimization, e.g., a hybrid differential evolution/genetic algorithm numerical optimization to design and optimize signal processing networks.

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 system (1) generates an output indicating scores for the extent of matching of pairs of data records. Thresholds may be set for the scores for decision-making or human review. A vector extraction module (12) measures similarity of pairs of fields in a pair of records to generate a vector. The vector is then processed to generate a score for the record pair.