Abstract: A method and a system for predicting a gas content in transformer oil based on a joint model are provided and belong a field of transformer failure prediction. The method includes the following: determining a type and a time series of gas to be predicted related to a failure, processing an original series by adopting empirical mode decomposition (EMD) and local mean decomposition (LMD) for a non-stationarity characteristic of a dissolved gas concentration series in oil; performing normalization on each sub-series component, dividing a training sample and a test sample; and establishing a deep belief network (DBN) prediction model for each of the sub-series components for training, performing superposition and reconstruction on the established DBN prediction model to perform characteristic extraction and classification on multi-dimensional data of the failure, evaluating prediction performance of the prediction model through calculating an error index.

Abstract: Autonomous vehicles are an exciting prospect to the future of driving. However, concerns about the decision-making made by the AI controlling a vehicle has been of concern, particularly in light of high-profile accidents. We can alleviate some concern, introduce better decisions, and also train an AI to make better decisions by introducing a remote viewer's, e.g., a human's, reaction to a possibly complex environment surrounding a vehicle that includes a potential threat to the vehicle. One or more remote viewer may provide a recommended response to the threat that may be incorporated in whole or in part in how the vehicle reacts. Various ways to engage and utilize remote viewers are proposed to improve the likelihood of receiving useful recommendations, including modifying how the environment is presented to a remote viewer to best suit the remote viewer, e.g., perhaps present the threat in a game.

Abstract: Method and system with federated learning model for health care applications are disclosed. The system for federated learning comprises multiple edge devices of end users, one or more federated learner update repository, and one or more cloud. Each edge device comprises a federated learner model, configured to send tensors to federated learner update repository. Cloud comprises a federated learner model, configured to send tensors to federated learner update repository. Federated learner update repository comprises a back-end configuration, configured to send model updates to edge devices and cloud.

Abstract: A hearing device system and the method for processing audio signals are described. The hearing device system has at least one hearing device having a recording device for recording an input signal, at least one neural network for separating at least one audio signal from the input signal and a playback device for playing back an output signal ascertained from the at least audio signal. A calibration device is connected to the at least one hearing device in the data-transmitting manner. The at least one neural network is customizable and/or replaceable by the calibration device.

Abstract: A memory of an electronic control unit stores a neural network into which travel state quantities of a vehicle are input and from which an internal temperature of a turbocharger is output. An arithmetic processing circuit of the electronic control unit executes an internal temperature estimation process of calculating, as an estimated value of the internal temperature, an output of the neural network using measured values of the travel state quantities as an input, and a deposit amount estimation process of calculating an amount of increase in an amount of oil coke deposits based on the estimated value of the internal temperature and calculating an estimated value of the amount of oil coke deposits as a value integrating the amount of increase.

Abstract: Examples described herein involve detecting known environmental conditions of using a neural network. An example implementation involves a playback device receiving data indicating a response of a listening environment to audio output of one or more playback devices as captured by a microphone and determining an input vector for a neural network. The playback device provides the determined input vector to the neural network, which includes an output layer comprising neurons that correspond to respective environmental conditions. The playback device detects that the input vector caused one or more neurons of the neural network to fire such that the neural network indicates that one or more particular environmental conditions are present in the listening environment. The playback device adjusts audio output of the one or more playback devices to at least partially offset the one or more particular environmental conditions.

Abstract: A computer-implemented method includes instantiating a neural network including a recurrent cell. The recurrent cell includes a probabilistic state component. The method further includes training the neural network with a sequence of data. In an embodiment, the method includes extracting a deterministic finite automaton from the trained recurrent neural network and classifying a sequence with the extracted automaton.

Abstract: A method and system for deep learning based cardiac electrophysiological model personalization is disclosed. Electrophysiological measurements of a patient, such as an ECG trace, are received. A computational cardiac electrophysiology model is personalized by calculating patient-specific values for a parameter of the computational cardiac electrophysiology model based at least on the electrophysiological measurements of the patient using a trained deep neural network (DNN). The parameter of the computational cardiac electrophysiology model corresponds to a spatially varying electrical cardiac tissue property.

Abstract: A system and method for multi-agent reinforcement learning with periodic parameter sharing that include inputting at least one occupancy grid to a convolutional neural network (CNN) and at least one vehicle dynamic parameter into a first fully connected layer and concatenating outputs of the CNN and the first fully connected layer. The system and method also include providing Q value estimates for agent actions based on processing of the concatenated outputs and choosing at least one autonomous action to be executed by at least one of: an ego agent and a target agent. The system and method further include processing a multi-agent policy that accounts for operation of the ego agent and the target agent with respect to one another within a multi-agent environment based on the at least one autonomous action to be executed by at least one of: the ego agent and the target agent.

Abstract: Systems and methods for training a residual neural network are described. One of the methods includes: initializing one or more weights of each of a plurality of residual units; inputting a plurality of training cases to a first warp unit in a series; using each training case to optimize the one or more weights for each residual unit in parallel in the first warp unit in the series; starting with the output of the first warp unit in the series, iteratively propagating the output of each warp unit to the input of a next respective warp unit in the series, for each respective warp unit, using each training case to optimize the one or more weights for each residual unit in parallel in the respective warp unit; and storing the output of the last warp unit in the series and the weights for each residual unit.

Abstract: Methods and apparatus, including computer program products, are provided for machine learning including deep convolutional neural networks. In some example embodiments, there may be provided a method that includes receiving, at a trained machine learning model, a portion of a test image; activating, at the machine learning model, a convolutional result formed based on the portion of the test image, the activation based on neighboring regions in the test image; and providing, by the machine learning model, an activated output. Related systems, methods, and articles of manufacture are also described.

Abstract: An uncertainty prediction apparatus includes an artificial neural network model trained based on deep learning, sampling models modeled by at least two weights obtained through sampling during a training process for the artificial neural network model, and an output generation unit configured to generate a result value reflecting an uncertainty degree by aggregating values output from the artificial neural network model and the sampling models after the same data is input to the artificial neural network model and the sampling models.

Abstract: Autonomous vehicles are an exciting prospect to the future of driving. However, concerns about the decision-making made by the AI controlling a vehicle has been of concern, particularly in light of high-profile accidents. We can alleviate some concern, introduce better decisions, and also train an AI to make better decisions by introducing a remote viewer's, e.g., a human's, reaction to a possibly complex environment surrounding a vehicle that includes a potential threat to the vehicle. One or more remote viewer may provide a recommended response to the threat that may be incorporated in whole or in part in how the vehicle reacts. Various ways to engage and utilize remote viewers are proposed to improve the likelihood of receiving useful recommendations, including modifying how the environment is presented to a remote viewer to best suit the remote viewer, e.g., perhaps present the threat in a game.

Abstract: The present disclosure provides systems, methods, and computer program products for controlling an object. An example method can comprise (a) obtaining video data of the object and (b) performing motion analysis on the video data to generate modified video data. The method can further comprise (c) using artificial intelligence (AI) to identify a set of features in the modified video data. The set of features may be indicative of a predicted state of the object. The AI may be been trained offline on historical training data. The method can further comprise (d) using the predicted state to determine a control signal and (e) transmitting, in real-time, the control signal to the object to adjust or maintain a state of the object in relation to the predicted state. Operations (a) to (d) can be performed without contacting the object.

Abstract: Few-shot learning of repetitive human tasks is performed. Sliding window-based temporal segmentation is performed of sensor data for a plurality of cycles of a repetitive task. Motion alignment is performed of the plurality of cycles, the motion alignment mapping portions of the plurality of cycles to corresponding portions of other of the plurality of cycles. Categories are constructed for each of the corresponding portions of the plurality of cycles according to the motion alignment. Meta-training is performed to teach a model according to data sampled from a labeled set of human motions and the categories for each of the corresponding portions, the model utilizing a bidirectional long short-term memory (LSTM) network to account for length variation between the plurality of cycles. The model is used to perform temporal segmentation on a data stream of sensor data in real time for predicting motion windows within the data stream.

Abstract: A method of converting one or more datasheets into a multi-player simulation-based game.

Abstract: An engine health diagnostic apparatus is provided for analysing health of a reciprocating internal combustion engine. The apparatus comprises feature generation circuitry for processing vibration sensor data received from a vibration sensor detecting vibration at a component of the reciprocating internal combustion engine 4 and generating a feature vector indicating multiple features of the sensor data. Processing circuitry processes the feature vector using a trained classification model which is defined by model parameters characterising a decision boundary of healthy operation learnt from a training set of feature vectors captured during healthy operation of the engine. The model generates an engine health indication providing a quantitative indication of deviation of the feature vector from the decision boundary of healthy operation.

Abstract: A machine learning device which learns a correlation between shipment inspection information obtained by inspecting an object in shipment thereof and operation alarm information issued during operation of the object, includes a state observation unit which observes the shipment inspection information and the operation alarm information; and a learning unit which generates a learning model based on the shipment inspection information and the operation alarm information observed by the state observation unit.

Abstract: An ultrasound imaging system includes a processor to associate an ultrasound image with one or more pictographs that correspond to the image. A processor provides ultrasound image data to one or more trained neural networks that classify the image data. The processor is programmed to select one or more pictographs that correspond to the classified image data for presentation to the operator. In some embodiments, the processor is programmed to use a neural network to determine if an ultrasound image captured during an examination corresponds to one or more required views for a particular type of examination.

Abstract: A system control method includes receiving a control task and randomly selecting a chromosome from an evolution pool according to the control task. The selected chromosome is decoded to obtain (N+1) ensemble policies, where the chromosome includes (N+1) gene fragments, and where N is a positive integer greater than or equal to 1. Each gene fragment can uniquely correspond to an ensemble policy, and each ensemble policy can uniquely correspond to a preset function. One ensemble policy is used for assigning a weight to a preset function that uniquely corresponds to the ensemble policy. The evolution pool can maintain two or more chromosomes. An ensemble calculation is performed according to weights assigned by the (N+1) ensemble policies to obtain an ensemble control output. A control signal is generated according to the ensemble control output, where the control signal is used for performing system control.

Abstract: An apparatus, a recording medium, and a method for generating a control parameter of a screen printer are disclosed. The apparatus includes a memory that stores a simulation model configured to derive predictive inspection information on a printed state of solder paste based on a plurality of control parameters of the screen printer; a communication circuit configured to receive first inspection information on a plurality of solder pastes printed by the screen printer based on a first control parameter, and a processor electrically connected to the memory and the communication circuit.

Abstract: A method for using an electrochemical energy storage device within a specifiable usage plan time period so as to optimize the service life is provided. The calendar aging constitutes a variable which influences the service life of the energy storage device and which is expressed in the form of a nonlinear function based on the fill state and the temperature of the energy storage device, the temperature being influencable by at least one charging and/or discharging process.

Abstract: An evaluation device may include: a receiving unit that receives an image of the semiconductor substrate, the image captured by an imaging device provided on the semiconductor substrate manufacturing apparatus; a determination unit that determines, using a neural network, at least one value representative of a probability of a machine learning device outputting an erroneous output for the image, the machine learning device configured to: (i) receive the image of the semiconductor substrate, (ii) perform computation using the received image, and (iii) output information indicating the quality of the semiconductor substrate based on a result of the computation; and an output unit that outputs an output based on the at least one value representative of the probability. The neural network has been trained using: images of manufactured semiconductor substrates; and information indicating, for each one of the images, a level of erroneousness for an output from the machine learning device.

Abstract: Systems and methods for transforming legacy models and transforming a model into a neural network model are disclosed. In an embodiment, a method may include receiving input data comprising an input model, an input dataset, and an input command. The method may include applying the input model to the input dataset to generate model output and storing model output and at least one of input model features or a map of the input model. The method may include generating a candidate neural network models with parameters. The method may include tuning the candidate neural network models to the input model. The method may include receiving model output from the candidate neural network models and selecting a neural network model from the candidate neural network models based on the candidate model output and the model selection criteria. In some aspects, the method may include returning the selected neural network model.

Abstract: A decision aid method for determining an action to be implemented by a given competitive entity in a competitive system comprises the competitive entity and at least one other adverse competitive entity, the competitive entity being able to implement an action from among a set of predefined actions, each action providing a different expected gain as a function of the actions implemented by the adverse competitive entities.

Abstract: A system, method, and computer-readable medium are disclosed for predicting a defect within a computer program comprising: accessing a code base of the computer program, the code base of the computer program comprising a plurality of computer program files; training the defect prediction system, the training including performing a historical analysis of defect occurrence patterns in the code base of the computer program; analyzing a commit of the computer program to identify a likelihood of defect occurrence within each of the plurality of files of the computer program; and, calculating a defect prediction metric for each of the plurality of files of the computer program, the defect prediction metric providing an objective measure of defect prediction for each of the plurality of files of the computer program.

Abstract: Techniques for performing Monte Carlo simulations within a provider network are described. A request to perform a Monte Carlo simulation is received that specifies a template identifying input parameters to the simulation. The simulation is performed in one or more virtual machines (VMs) by sampling input values from an input data distribution stored at a first network location specified in the template and by processing the sampled input values using an input parameter specified in the template to form an output distribution that is stored at a second network location specified in the template. A statistic of the output distribution is generated and sent to an electronic device.

Abstract: Various systems and methods are described for correlating technology choices with the risk of system vulnerabilities. A system captures and quantifies both observations of technology choices as well as the outputs certain outputs of internal choices and processes across a number of different organizations. A Bayesian estimate of vulnerability is imputed from the choices and observed use of vulnerable technology, further segmented by business type, revenue, and size. Differences between the observation of a particular organization and Bayesian expected value are measured and converted to vulnerability score, the vulnerability score embodying a point-in-time and longitudinal measure of organizational performance, including the likelihood of future compromise due to software vulnerabilities. The vulnerability score can then be further used to price risk, for example in a cyber insurance context.

Abstract: Described herein is a system transmits and combines local models, that individually include a set of local parameters computed via stochastic gradient descent (SGD), into a global model that includes a set of global model parameters. The local models are computed in parallel at different geographic locations (e.g., different instances of computing infrastructure) along with symbolic representations. Network transmission of the local models and the symbolic representations, rather than transmission of the large training data subsets processed to compute the local models and symbolic representations, conserves resources and decreases latency. The global model can then be used as a model to determine a likelihood that at least a portion of current and/or recently received data traffic is illegitimate data traffic that is associated with a cyber attack. In some instances, the system can implement a remedial action to mitigate the effects of the cyber attack on computing infrastructure.

Abstract: A method and system of provisioning a release of an application are provided. An average arrival rate for the application is calculated. A residual capacity of each node in a pool of nodes is determined. A risk dependent state of each node is determined. For each node, a failure index based on the average arrival rate of the application, the residual capacity of the node, and the risk dependent state of the node, is calculated. The release of the application is provisioned on a node in the pool of nodes, based on the calculated failure index.

Abstract: Systems, methods, and devices of the various embodiments provide a hardware and software architecture enabling electrochemical impedance spectroscopy (“EIS”) to be performed on multiple electrochemical devices, such as fuel cells, at the same time without human interaction with the electrochemical devices and to use EIS to dynamically monitor the performance of a fuel cell system. Embodiment methods may include determining an impedance of a set of fuel cells using electrochemical impedance spectroscopy, determining an ohmic polarization of the set of fuel cells from the impedance, determining a concentration polarization of the set of fuel cells from the impedance, comparing the ohmic polarization of the set of fuel cells to a first threshold, comparing the concentration polarization of the set of fuel cells to a second threshold, and initiating a corrective action when the ohmic polarization is above the first threshold or when the concentration polarization is below the second threshold.

Abstract: Resolution enhancement techniques are described. An apparatus may receive first image data at a first resolution, and second image data at a resolution less than the first resolution. The second image data may be scaled to the first resolution and compared to the first image data. Application of a neural network may scale the first image data to a resolution higher than the first resolution. The application of the neural network may incorporate signals based on the scaled second image data. The signals may include information obtained by comparing the scaled second image data to the resolution of the first image data.

Abstract: The present disclosure is related to a battery management system which includes a location information obtainer configured to obtain location information of a battery, and an estimation model changer configured to change an estimation model to estimate an internal state of the battery according to a change in the location information.

Abstract: The invention relates to a method for determining critical operating states in a fuel cell stack, consisting of single cells connected in series, wherein a low-frequency current or voltage signal is applied to the fuel cell stack, the resulting voltage or current signal is measured and the distortion factor thd is determined. According to the invention, the weighted sum of a term dependent on the membrane resistance RM and a term dependent on the distortion factor thd is used to determine an indicator THDAdryout correlating with the drying out of the fuel cell membranes of the fuel cell stack, the membrane resistance Rm being detected by impedance measurement.

Abstract: A method and an apparatus for providing a safe operation of a technical system including a plurality of system components. The method includes the steps of: a) providing a safety analysis model matured by knowledge about former implementations of the respective system components in different context, b) whereby system components' dependencies are modeled by connecting inports with outports of the respective system components and/or vice versa, c) whereby at least one or a plurality of such in and/or outports are associated with input failure modes and/or output failure modes, d) characterized in automatically uncovering inconsistencies caused by at least one system component to be integrated in connection with at least another system component whereby the input and/or output failure mode of the system component carries the knowledge from another implementation into the context.

Abstract: Some embodiments described herein cover a machine learning architecture with a separated perception subsystem and application subsystem. These subsystems can be co-trained. In one example embodiment, a data item is received and information from the data item is processed by a first node to generate a first feature vector comprising a plurality of features, each of the plurality of features having a similarity value representing a similarity to one of a plurality of centroids. The first node selects a subset of the features from the first feature vector, the subset containing one or more features that have highest similarity values. The first node generates a second feature vector from the first feature vector by replacing similarity values of features in the first feature vector that are not in the subset with zeros. A second node then processes the second feature vector to determine an output.

Abstract: An apparatus, method, and computer program that can learn a linear or non-linear transfer function are disclosed. In one embodiment, for example, a computer-implemented method may include creating a plurality of combinations for at least one input, and calculating an output for the plurality of combinations. The method may also include calculating an error based on the calculated output.

Abstract: Computer implemented systems and methods are disclosed for structuring data from unstructured electronic data files. In accordance with some embodiments, an electronic data file including unstructured content associated with a legal process return is received and the unstructured content parsed. The unstructured content is parsed to identify one or more objects and properties based on a database ontology that are processed to generate an object model. A data report may be generated based on the identified objects and properties.

Abstract: Co-occurrence data representing e.g. preferences and facts observed in a plurality of situations may be stored in a matrix as combinations of high-dimensional sparse vectors. The matrix may be called e.g. as an experience matrix. The data stored in the experience matrix may be subsequently utilized e.g. for predicting a preference of a user in a new situation. A prediction may be determined by a method comprising providing a query comprising one or more query words, accessing the experience matrix containing co-occurrence data stored as vectors of the experience matrix, determining a first auxiliary vector by identifying a vector of the experience matrix associated with a first query word, forming a query vector by using the first auxiliary vector, and determining the prediction by comparing the query vector with the vectors of the experience matrix.

Abstract: A gait perturbation system and a method for testing and/or training using the gait perturbation system is disclosed herein. The gait perturbation system includes a gait perturbation device and a data processing device. The gait perturbation device includes one or more displaceable components configured to be displaced at a plurality of different speeds, and having one or more respective surfaces for receiving one or more respective limbs of a person; and one or more speed adjustment mechanisms coupled to the one or more displaceable components to adjust the speed set point at which the one or more displaceable components are displaced. The data processing device is configured to generate a signal for introducing a perturbation to the one or more displaceable components, and to control the speed set point of the one or more displaceable components such that the one or more displaceable components perturb a gait of the person.

Abstract: Method embodiments are provided herein for dynamically calibrating and adjusting a direct conversion receiver system. One embodiment includes applying one or more gain control signals to one or more gain elements of a receiver system, where the applying one or more gain control signals results in a gain change to the receiver system; in response to the gain change, determining whether the receiver system exhibits a DC (direct conversion) offset; and in response to a determination that the receiver system exhibits the DC offset, applying one or more DC offset correction control signals to one or more gain elements of the receiver system, where the one or more DC offset correction signals are configured to correct the DC offset.

Abstract: An adaptive fuzzy rule controlling system for a software defined storage (SDS) system to control performance parameters in a storage node is disclosed. The system includes: a traffic monitoring module, for acquiring observed values of performance parameters in the storage node; an adaptive neural fuzzy inference module, for learning a dynamic relationship between configurations of a plurality of storage devices in the storage node and the performance parameters during a period of time, and outputting fuzzy rules which is built according to the dynamic relationship; a traffic forecasting module, for providing forecasted values of the performance parameters in a particular point in time in the future; and a fuzzy rule control module, for arranging the configuration of the storage devices in the storage node in the particular point in time in the future according to the fuzzy rules and the forecasted values.

Abstract: A method of assessing a riparian area includes mapping a physical extent of a potential riparian area (PRA) along a stream using first imagery, determining a percent of basic functional groups of vegetation with the PRA using the first imagery, and determining a Normalized Difference Vegetation Index (NDVI) for the PRA over multiple years using second imagery. The method further includes setting an NDVI threshold value for riparian vegetation in the PRA using the NDVI, mathematically relating the percent of basic functional groups with the NDVI threshold value to evaluate and correct the NDVI threshold value, and determining a proper function and condition (PFC) assessment of the PRA using the corrected NDVI threshold value.

Abstract: A context-awareness system has an event data producer for forming event data from node data including a node value and node information, a context producer, and an event data manager for storing event data formation suppression information and event data conversion rules. The context producer sends the number of entries of event data (the number of inputs) to the event data manager per unit time. The manager sends to the event data producer the event data formation suppression information and event data conversion rules corresponding to the number of inputs. The event data producer forms event data depending on the event data formation suppression information and event data conversion rules corresponding to the number of inputs. Appropriate context information is formed without increasing the amount of resources when data including values related to events are entered in large amount and frequently.

Abstract: Disclosed is a clinical diagnosis supporting method and system based on hierarchical fuzzy inference. The clinical diagnosis supporting method includes the steps of: (a) extracting check data of a plurality of check items for each of a plurality of patients from a respective check item database; (b) selecting a characteristic from the check data of each of the check items; (c) hierarchical fuzzy-inferring the check data from which the characteristic is selected, for each check item; (d) extracting a check opinion for each disease based on an output value of the fuzzy inference for each check item; and (e) making a synthetic clinical estimation based on the extracted check opinion.

Abstract: A method of controlling a fuel cell system includes applying alternating current (AC) signals to an individual fuel cell. The AC signals have a plurality of different frequencies. A voltage across the individual fuel cell is determined at each of the plurality of different frequencies. An impedance characteristic of the individual fuel cell is determined based at least in part on the voltage across the individual fuel cell at each of the plurality of different frequencies. The individual fuel cell is controlled based at least in part on the impedance characteristic.

Abstract: An image processing device includes: an interest area detector that detects interest areas included in a time-series image group captured in time series; a calculation processing unit that calculates feature amounts indicative of features of the interest areas; an area classification unit that classifies the interest areas into area groups, based on the feature amounts of the interest areas and time-series positions of time-series images including the interest areas; a group feature amount calculation unit that calculates a group feature amount indicative of a feature of each of the area groups; an area selection unit that selects one or more representative areas of the interest areas belonging to the area groups, from among the area groups; and a representative image output unit that outputs one or more representative images including the representative areas in the time-series image group.

Abstract: An information processing apparatus includes a network learning portion that performs learning of an appearance/position recognition network by constraining first to third weights and using a learning image, wherein the appearance/position recognition network has a foreground layer including a position node, a background layer including a background node, and an image layer including a pixel node, and is a neural network in which the position node, the background node, and the pixel node are connected to each other, and wherein the first weight is a connection weight between the position node and the pixel node, the second weight is a connection weight between the position node and the background node, and the third weight is a connection weight between the background node and the pixel node.

Abstract: A system and method of automatic local grid refinements in reservoir simulation systems is described. In one aspect of the disclosure, a method is directed to a computer-implemented method of modeling a formation. The method includes applying a coarse grid to a geologic formation of interest, the coarse grid having a plurality of coarse grid blocks and identifying a structure of interest disposed in the formation. Further, the method includes determining a fine grid zone around the structure based upon a time period for flow simulation of the geologic formation and a geologic characteristic of the geologic formation in a local region adjacent the structure and applying a fine grid to the coarse grid blocks encompassed by the fine grid zone.

Abstract: A neural network process for improving wind retrievals from wind profiler measurements is described. In this invention, a neural network is trained to retrieve (missing or incomplete) upper level winds from ground based wind profiler measurements. Radiosonde measurements in conjunction with wind profiler ground measurements for specific geographical locations are used as training sets for the neural network. The idea is to retrieve timely and spatially continuous upper level wind information from (fragmented or incomplete) wind profiler measurements.