Abstract: Implementations relate to using deep reinforcement learning to train a model that can be utilized, at each of a plurality of time steps, to determine a corresponding robotic action for completing a robotic task. Implementations additionally or alternatively relate to utilization of such a model in controlling a robot. The robotic action determined at a given time step utilizing such a model can be based on: current sensor data associated with the robot for the given time step, and free-form natural language input provided by a user. The free-form natural language input can direct the robot to accomplish a particular task, optionally with reference to one or more intermediary steps for accomplishing the particular task. For example, the free-form natural language input can direct the robot to navigate to a particular landmark, with reference to one or more intermediary landmarks to be encountered in navigating to the particular landmark.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing a compressed recurrent neural network (RNN). One of the systems includes a compressed RNN, the compressed RNN comprising a plurality of recurrent layers, wherein each of the recurrent layers has a respective recurrent weight matrix and a respective inter-layer weight matrix, and wherein at least one of recurrent layers is compressed such that a respective recurrent weight matrix of the compressed layer is defined by a first compressed weight matrix and a projection matrix and a respective inter-layer weight matrix of the compressed layer is defined by a second compressed weight matrix and the projection matrix.

Abstract: Implementations relate to using deep reinforcement learning to train a model that can be utilized, at each of a plurality of time steps, to determine a corresponding robotic action for completing a robotic task. Implementations additionally or alternatively relate to utilization of such a model in controlling a robot. The robotic action determined at a given time step utilizing such a model can be based on: current sensor data associated with the robot for the given time step, and free-form natural language input provided by a user. The free-form natural language input can direct the robot to accomplish a particular task, optionally with reference to one or more intermediary steps for accomplishing the particular task. For example, the free-form natural language input can direct the robot to navigate to a particular landmark, with reference to one or more intermediary landmarks to be encountered in navigating to the particular landmark.

Abstract: A method for a radar device is described below. According to an example implementation, the method comprises transmitting an RF transmission signal that comprises a plurality of frequency-modulated chirps, and receiving an RF radar signal and generating a dataset containing in each case a particular number of digital values based on the received RF radar signal. A dataset may in this case be associated with a chirp or a sequence of successive chirps. The method furthermore comprises filtering the dataset by way of a neural network to which the dataset is fed in order to reduce an interfering signal contained therein. A convolutional neural network is used as the neural network.

Abstract: An operation device includes a quantizer circuit, a buffer circuit, a convolution core circuit and a multiply-add circuit. The quantizer circuit receives first feature data and performs asymmetric uniform quantization on the first feature data to obtain and store in the buffer circuit second feature data. The quantizer circuit further receives a first weighting coefficient and performs symmetric uniform quantization on the first weighting coefficient to obtain and store in the buffer circuit a second weight coefficient. The convolution core circuit performs a convolution operation on the initial operation result, an actual quantization scale factor and an actual bias value to obtain a final operation result.

Abstract: Systems, methods, and devices for training and testing utterance based frameworks are disclosed. The training and testing can be conducting using synthetic utterance samples in addition to natural utterance samples. The synthetic utterance samples can be generated based on a vector space representation of natural utterances. In one method, a synthetic weight vector associated with a vector space is generated. An average representation of the vector space is added to the synthetic weight vector to form a synthetic feature vector. The synthetic feature vector is used to generate a synthetic voice sample. The synthetic voice sample is provided to the utterance-based framework as at least one of a testing or training sample.

Abstract: A bio-signal data processing apparatus includes a communicator configured to receive electrocardiogram data from a bio-signal measuring apparatus, a recording unit configured to record the electrocardiogram data, a transmission delay determiner, and an output information generator. The transmission delay determiner is configured to generate transmission delay information by comparing a recording time of the electrocardiogram data with a reception time of the electrocardiogram data, detect whether or not a delay according to data transmission occurs, by considering the transmission delay information, and, when the delay is detected to occur, calculating delay time information that is calculated on the basis of the transmission delay information. The output information generator is configured to correct the electrocardiogram data by using the delay time information and generate output data of the electrocardiogram data corresponding to a user input.

Abstract: A neural network includes a drop layer configured to drops feature values. A method of computation using the neural network includes extracting feature data from input data using a first portion of a neural network, generating compressed representation data of the extracted feature data by dropping a feature value from the extracted feature data at a drop layer of the neural network based on a drop probability corresponding to the feature value, and indicating an inference result from the compressed representation data using a second portion of the neural network.

Abstract: A method of establishing an image set for image recognition includes obtaining a single-label image set comprising an image annotated with a single label, and a multi-label image set comprising an image annotated with a plurality of labels; converting content of each label into a corresponding word identifier according to a semantic network, to obtain a word identifier set, a converted single-label image set, and a converted multi-label image set; and constructing a hierarchical semantic structure according to the word identifier set and the semantic network. The method also includes performing label supplementation on the image in the converted single-label image set to obtain a supplemented single-label image set; performing label supplementation on the supplemented single-label image set to obtain a final supplemented image set; and establishing a target multi-label image set to train an image recognition model by using the target multi-label image set.

Abstract: An output time-series of a cell of a neural network is captured. A subset of a set of data points of the output time-series is consolidated into a singular data point. The singular data point is fitted in a data representation to form a quantified aggregated data point. The quantified aggregated data point is included in an intermediate time-series. Using the intermediate time-series as an input at an intermediate layer of the neural network, an anonymized output time-series is produced from the neural network.

Abstract: An embodiment according to the present invention includes a method for a machine-learning based approach to the formation of ultrasound and photoacoustic images. The machine-learning approach is used to reduce or remove artifacts to create a new type of high-contrast, high-resolution, artifact-free image. The method of the present invention uses convolutional neural networks (CNNs) to determine target locations to replace the geometry-based beamforming that is currently used. The approach is extendable to any application where beamforming is required, such as radar or seismography.

Abstract: A filter design method for a small target detection on infrared imagery using a normalized-cross-correlation layer in neural networks, including the steps of: Normalizing inputs and filters of a convolutional neural network, wherein normalizing inputs and filters of the convolutional neural network provides faster convergence in a limited database. Defining a forward function of a normalization layer in the convolutional neural network, wherein the forward function of the normalization layer in the convolutional neural network is used for training a neural network. Defining a derivative function of the normalization layer for a back propagation in a neural network training phase. Training created neural networks with datasets, wherein the datasets consist of target and background views and using trained neural networks in the small target detection.

Abstract: Disclosed is an operation method of an audio signal processing device configured to process an audio signal including a first audio signal component and a second audio signal component. The operation method includes: receiving the audio signal; normalizing loudness of the audio signal, based on a pre-designated target loudness; acquiring the first audio signal component from the audio signal having the normalized loudness, by using a machine learning model; and de-normalizing loudness of the first audio signal component, based on the pre-designated target loudness.

Abstract: Devices and techniques are generally described for per-image printer setting optimization. In some examples, first data representing first input image data may be generated. A classifier network may generate first category data representing a predicted image category for the first input image data based at least in part on the first data. Second data representing a color of a printing medium and third data associated with a material of the printing medium may be determined. First printer configuration data may be determined by searching a first data structure using a combination of the first category data, the second data, and the third data. The first input image data and the first printer configuration data may be sent to the first printer. The first printer may be effective to print the first input image data using settings specified by the first printer configuration data.

Abstract: Embodiments of the present disclosure relate to a data processing method and apparatus for a neural network. The neural network is provided with at least one activation function. A method may include: converting, in response to that an activation function acquiring current data is a target function, based on a conversion relationship between the target function and a preset function, the current data into input data of the preset function; finding out first output data of the preset function with the input data as an input in a lookup table corresponding to the preset function; obtaining second output data of the target function with the current data as an input by conversion based on the conversion relationship and the first output data; and outputting the second output data.

Abstract: Validating electronic content by users includes providing, by a data processing system, electronic content to users, receiving, by the data processing system, the ratings of the electronic content by the at least some of the users based on the rating metric(s), each rating being a raw rating having a default weight, adjusting, by the data processing system, the default weight for each raw rating based on one or more of pre-rating user behaviors, a user rating history and a user credibility rating to arrive at an adjusted rating, using, by the data processing system, the adjusted rating to arrive at a total rating, and providing the total rating to the users.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for clustering media items in a semantic space to generate theme-based folders that organize media items by content theme. In particular, the disclosed systems can access media items that are stored in an original folder structure. The disclosed systems can generate content-based tags for each media item in a collection of media items. Based on the generated tags, the disclosed systems can map the collection of media items to a semantic space and cluster the collection of media items. The disclosed systems determine themes for the clusters based on the generated tags. The disclosed systems can present a media item navigation graphical user interface comprising the collection of media items organized by themes. The disclosed system can present the media item navigation graphical user interface without altering the original folder structure.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a projection neural network. In one aspect, a projection neural network is configured to receive a projection network input and to generate a projection network output from the projection network input. The projection neural network includes a sequence of one or more projection layers. Each projection layer has multiple projection layer parameters, and is configured to receive a layer input, apply multiple projection layer functions to the layer input, and generate a layer output by applying the projection layer parameters for the projection layer to the projection function outputs.

Abstract: Systems and methods for predicting performance of a modulation system are provided. A neural network model is trained using performance information of a source system. The neural network model is modified with transferable knowledge about a target system to be evaluated. The neural network model is tuned using specific characteristics of the target system to create a source-based target model. The target system performance is evaluated using the source-based target model to predict system performance of the target system.

Abstract: A system and method that provides balanced pruning of weights of a deep neural network (DNN) in which weights of the DNN are partitioned into a plurality of groups, a count of a number of non-zero weights is determined in each group, a variance of the count of weights in each group is determined, a loss function of the DNN is minimized using Lagrange multipliers with a constraint that the variance of the count of weights in each group is equal to 0, and the weights and the Lagrange multipliers are retrained by back-propagation.

Abstract: Systems and methods for generating regressors based on data sparsity using a machine learning (ML) model are described. A system is configured to provide a plurality of time series datasets to a recurrent neural network (RNN) of a machine learning (ML) model. The RNN generates one or more outputs associated with one or more time series datasets, and the system provides a first portion and a second portion of the one or more outputs to a regressor layer and a classification layer of the ML model, respectively. The regressor layer generates one or more regressors for the one or more time series datasets, and the classification layer generates one or more classifications associated with the one or more regressors (with each indicating whether an associated regressor is valid). Whether a classification indicates a regressor is valid may be based on time series data sparsity.

Abstract: A computer system is provided for converting images through use of a trained neural network. A source image is divided into blocks and context data is added to each pixel block. The context blocks are split into channels and each channel from the same context block is added to the same activation matrix. The action matrix is then executed against a trained neural network to produce a changed activation matrix. The changed activation matrix is then used to generate a converted image.

Abstract: This disclosure relates generally to authenticating humans based on behavioral pattern. The method and system proposed provides a continuous/seamless monitoring platform for authenticating humans by continuously monitoring routine activities of subjects (Activities of Daily Living (ADL)) in a smart environment using plurality of passive, unobtrusive, binary, unobtrusive non-intrusive sensors embedded in living infrastructure. The proposed method and system for authenticating humans based on behavioral pattern is provided. The daily routine activities of humans/subjects, housed in a smart environment is continuous monitored by plurality of non-intrusive sensors embedded in living infrastructure. Further the collected sensor data is processed in several stages, which includes pre-processing of sensor data, behavioral pattern prediction, error detection based on predicted behavioral pattern and so on for authenticating humans based on behavioral pattern.

Abstract: The present disclosure provides an end-to-end model training method and apparatus, which relates to a field of artificial intelligence technologies. The method includes: obtaining training data containing a plurality of training samples, in which the plurality of training samples include an original sequence, a target sequence and a corresponding tag list, the tag list includes importance tags in the target sequence and avoidance tags corresponding to the importance tags, and the avoidance tags are irrelevant tags corresponding to the importance tags; and adopting the training data to train a preset end-to-end model until a value of a preset optimization target function is smaller than a preset threshold.

Abstract: A method and a system for implementing neural network models on edge devices in an Internet of Things (IoT) network are disclosed. In an embodiment, the method may include receiving a neural network model trained and configured to detect objects from images, and iteratively assigning a new value to each of a plurality of parameters associated with the neural network model to generate a re-configured neural network model in each iteration. The method may further include deploying for a current iteration the re-configured neural network on the edge device. The method may further include computing for the current iteration, a trade-off value based on a detection accuracy associated with the at least one object detected in the image and resource utilization data associated with the edge device, and selecting the re-configured neural network model, based on the trade-off value calculated for the current iteration.

Abstract: A machining environment estimation device includes a data acquisition unit that acquires vibration time-series data which indicates a machining environment of the machine tool, machining conditions in carrying out machining for a workpiece in the machine tool, measurement data of a machined surface of a machining-finished workpiece, and machined surface evaluation data, a pre-processing unit that creates vibration data and machining condition data which serve as state data, and machined surface measurement data and machined surface evaluation data which serve as label data, and a learning unit that generates a learning model which learned (a) the machined surface measurement data and (b) a machined surface evaluation result of the machining-finished workpiece, with respect to (i) a vibration state and (ii) the machining conditions in the machining environment, based on the state data and the label data.

Abstract: In some implementations, the present disclosure relates to a method. The method includes obtaining a set of weights for a neural network comprising a plurality of nodes and a plurality of connections between the plurality of nodes. The method also includes identifying a first subset of weights and a second subset of weights based on the set of weights. The first subset of weights comprises weights that used by the neural network. The second subset of weights comprises weights that are prunable. The method further includes storing the first subset of weights in a first portion of a memory. A first error correction code is used for the first portion of the memory. The method further includes storing the second subset of weights in a second portion of the memory. A second error correction code is used for the second portion of the memory. The second error correction code is weaker than the first error correction code.

Abstract: The present disclosure relates to systems and methods for detecting and identifying anomalies within a discrete wavelet database. In one implementation, the system may include one or more memories storing instructions and one or more processors configured to execute the instructions. The instructions may include instructions to receive a new wavelet, convert the net transaction to a wavelet, convert the wavelet to a tensor using an exponential smoothing average, calculate a difference field between the tensor and a field having one or more previous transactions represented as tensors, perform a weighted summation of the difference field to produce a difference vector, apply one or more models to the difference vector to determine a likelihood of the new wavelet representing an anomaly, and add the new wavelet to the field when the likelihood is below a threshold.

Abstract: A system, method and computer product for training a neural network system. The method comprises inputting an audio signal to the system to generate plural outputs f(X, ?). The audio signal includes one or more of vocal content and/or musical instrument content, and each output f(X, ?) corresponds to a respective one of the different content types. The method also comprises comparing individual outputs f(X, ?) of the neural network system to corresponding target signals. For each compared output f(X, ?), at least one parameter of the system is adjusted to reduce a result of the comparing performed for the output f(X, ?), to train the system to estimate the different content types. In one example embodiment, the system comprises a U-Net architecture. After training, the system can estimate various different types of vocal and/or instrument components of an audio signal, depending on which type of component(s) the system is trained to estimate.

Abstract: The present invention is directed to an apparatus for minimizing power consumption in a cooling system. In one embodiment, the apparatus comprises one or more processors, one or more sensors associated with one or more regulated environments and one or more chillers that regulate temperature of the one or more regulated environments and a storage device, coupled to the one or more processors, storing instructions that when executed by the one or more processors performs a method. The method comprises gathering readings from the one or more sensors, determining a cost and power consumption associated with setting values for a plurality of control variables associated with the one or more chiller plants, selecting values for the control variables with a minimum cost as optimized control variable values and applying the optimized control variable values to the plurality of control variables to minimize power consumption of the cooling system.

Abstract: In a method of operating a neural network device, a plurality of consecutive input data is received by an input layer. A delta data is generated by the input layer based on a difference between a current input data and a previous input data. A first current feature is generated by a first linear layer based on a first delta feature generated by performing a first linear operation on the delta data and a first previous feature. A second delta feature is generated by a first nonlinear layer based on a second current feature generated by performing a first nonlinear operation on the first current feature and a second previous feature. A third current feature is generated by a second linear layer based on a third delta feature generated by performing a second linear operation on the second delta feature and a third previous feature.

Abstract: Methods for determining a fixed point format for one or more layers of a DNN based on the portion of the output error of the DNN attributed to the fixed point formats of the different layers. Specifically, in the methods described herein the output error of a DNN attributable to the quantisation of the weights or input data values of each layer is determined using a Taylor approximation and the fixed point number format of one or more layers is adjusted based on the attribution. For example, where the fixed point number formats used by a DNN comprises an exponent and a mantissa bit length, the mantissa bit length of the layer allocated the lowest portion of the output error may be reduced, or the mantissa bit length of the layer allocated the highest portion of the output error may be increased. Such a method may be iteratively repeated to determine an optimum set of fixed point number formats for the layers of a DNN.

Abstract: Provided are an apparatus and method for linearly approximating a deep neural network (DNN) model which is a non-linear function. In general, a DNN model shows good performance in generation or classification tasks. However, the DNN fundamentally has non-linear characteristics, and therefore it is difficult to interpret how a result from inputs given to a black box model has been derived. To solve this problem, linear approximation of a DNN is proposed. The method for linearly approximating a DNN model includes 1) converting a neuron constituting a DNN into a polynomial, and 2) classifying the obtained polynomial as a polynomial of input signals and a polynomial of weights.

Abstract: A GNSS receiver for generating distance estimates from multiple GNSS satellites. The GNSS receiver includes an antenna and an RF front end coupled to the antenna configured to generate a plurality of samples related to a received signal. The GNSS receiver includes a correlator coupled to the RF front end configured to perform various operations including performing three correlations on the plurality of samples with three local code to generate three correlation results, where the three local codes are shifted in time or distance with respect to each other. The GNSS receiver includes a processor for defining a first slope using the first correlation result and the second correlation result, defining a second slope using the second correlation result and the third correlation result, and defining a code discriminator as a sum of the first slope and the second slope.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing an input sequence using a recurrent neural network to generate an output for the input sequence. One of the methods includes receiving the input sequence; generating a doubled sequence comprising a first instance of the input sequence followed by a second instance of the input sequence; and processing the doubled sequence using the recurrent neural network to generate the output for the input sequence.

Abstract: An apparatus and method for monitoring the productivity of a portable machine are provided. The method includes receiving motion data for at least one component of the portable machine from a multi-axis accelerometer, receiving position data for the at least one component from a process parameter sensor communicatively coupled to the at least one component, and determining, based on the received motion data and the received position data that the at least one component is oriented in a predetermined position for productive operation. The method also includes determining an area of productive operation using at least one physical dimension of the at least one component and the received motion data when the at least one component is oriented in the predetermined position for productive operation and incrementing a total area counter based on the determination.

Abstract: A method includes receiving N pairs of training examples and class labels therefor. Each pair includes a respective anchor example, and a respective non-anchor example capable of being a positive or a negative training example. The method further includes extracting features of the pairs by applying a DHCNN, and calculating, for each pair based on the features, a respective similarly measure between the respective anchor and no example. The method additionally includes calculating a similarity score based on the respective similarity measure for each pair. The score represents similarities between all anchor points and positive training examples in the pairs relative to similarities between all anchor points and negative training examples in the pairs. The method further includes maximizing the similarity score for the anchor example for each pair to pull together the training examples from a same class while pushing apart the training examples from different classes.

Abstract: The present invention relates broadly to a method of digitally filtering a signal, such as an audio signal, using a digital filter. The digital filter includes a plurality of neighbouring sample points broken into zones having different frequency content or frequency ranges. The zones adjacent one another may have neighbouring sample points in common. Generally each zone has at least same distinct frequencies compared with other zones. That is, the zones are roughly dependent on the frequency content. The invention in its preferred form involves combining values for two or more of the neighbouring sample points for select of the zones depending on its frequency content. The values are combined so as to provide a modified zone having substantially the same number of sample points as the select zone. The modified zones together provide a modified filter to be applied to the signal.

Abstract: A system for teaching compositionality to convolutional neural networks includes an unmasked convolutional neural network comprising a first set of convolutional neural network layers; a first masked convolutional neural network comprising a second set of convolutional neural network layers; the unmasked convolutional neural network and the first masked convolutional network sharing convolutional neural network weights; the system training the unmasked and first masked convolutional neural networks simultaneously based on an objective function that seeks to reduce both discriminative loss and compositional loss.

Abstract: A system employing a plurality of brain/body-generated inputs to control multi-action operation includes a controllable device, that performs at least two actions via remote control, and a head-mounted user interface device. The UI device includes a user cranium-mounted frame, first and second sensors supported by the frame, a processor, and a transmitter. The first sensor includes an electrode for contacting the cranium when the frame is worn and adapted to receive a bioelectric signal from the cranium. The second sensor receives hands-free brain/body input from the user. The processor uses application software process input provided by the first and second sensors and generates different outputs, corresponding to different commands. The transmitter transmits signals, based on the different commands, to the controllable device to initiate the different actions performed by the device.

Abstract: The present invention provides a cloud service packet redirection method and system, and a cloud gateway, the method performed by the cloud gateway includes, if the cloud gateway determines that a DNS packet that is forwarded by a router in a redirection manner is a cloud service-related DNS packet, a record is maintained in a cloud IP table of the cloud gateway according to the DNS packet. Policy route configuration information is sent to the router according to the record maintained in the cloud IP table to instruct the router to maintain a policy route. The policy route instructs the router to redirect, to the cloud gateway, a cloud service packet that is indicated by the DNS packet.

Abstract: A method of selecting and presenting content based on learned user preferences is provided. The method includes providing a content system including a set of content items organized by genre characterizing the content items, and wherein the set of content items contains microgenre metadata further characterizing the content items. The method also includes receiving search input from the user for identifying desired content items and, in response, presenting a subset of content items to the user. The method further includes receiving content item selection actions from the user and analyzing the microgenre metadata within the selected content items to learn the preferred microgenres of the user. The method includes, in response to receiving subsequent user search input, selecting and presenting content items in an order that portrays as relatively more relevant those content items containing microgenre metadata that more closely match the learned microgenre preferences of the user.

Abstract: A synergy-based human-machine interface that uses low-dimensional command signals to control a high dimensional virtual, robotic or paralyzed human hand is provided. Temporal postural synergies are extracted from angular velocities of finger joints of five healthy subjects when they perform hand movements that are similar to activities of daily living. Extracted Synergies are used in real-time brain control, where a virtual, robotic or paralyzed human hand is controlled to manipulate virtual or real world objects.

Abstract: A system and method for filtering an already obtained information resource (“document”) for display is described, including identifying a target portion of the information resource; dividing the data of the target portion into elements of a set; receiving one or more search requests as a filter specification; interpreting the filter specification as a Boolean logic expression; evaluating the elements of the set based on the Boolean logic expression, in an ongoing manner, each time the filter specification is entered or modified; and dynamically re-rendering the display, upon each such entry or modification of the filter specification, with all elements not matching the logic expression removed from the display of the target portion, providing the user with a “shrinking document” more likely to contain just the information being looked for.

Abstract: Identifying a questionable network address from a network communication. In an embodiment, a network device receives an incoming or outgoing connection request, a web page, an email, or other network communication. An evaluation module evaluates the network communication for a corresponding network address, which may be for the source or destination of the network communication. The network address generally includes an IP address. The evaluation module determines one or more properties of the network communication, such as time of day, content type, directionality, or the like. The evaluation module then determines whether the properties match or are otherwise allowed based on properties specified in the white list in association with the IP address.

Abstract: Changes of a user's emotion that occurs when content is reproduced and a user's surrounding environment that occurs when the content is reproduced are recorded as sensing meta data and content is reproduced in a reproducing mode corresponding to the recorded sensing meta data. Content supplied from a content providing section 11 is normally reproduced. A human body sensor 24 measures biological information of a user of content. An environmental sensor 25 measures a user's surrounding environment. A sensing meta data generating section 19 generates sensing meta data using information detected by at least one of the human body sensor 24 and the environmental sensor 25. A recording processing section 20 records the generated sensing meta data to a record medium 21. A reproducing processing section 23 reproduces sensing meta data. A reproducing control section 12 reproduces content in a reproducing mode that changes corresponding to the reproduced sensing meta data.

Abstract: A method for learning connections between nonlinear oscillators in a neural network comprising the steps of providing a plurality of nonlinear oscillators, with each respective oscillator producing an oscillation distinct from the others in response to an input and detecting an input at an at least first oscillator of the plurality of nonlinear oscillators. Detecting an input at an at least a second oscillator of the plurality of nonlinear oscillators, comparing the oscillation of the at least first oscillator to the oscillation of the at least second oscillator at a point in time, and determining whether there is coherency between the oscillation of the at least first oscillator and the oscillation of the at least second oscillator. Changing at least one of the amplitude and phase of a connection between the at least first oscillator and the at least second least oscillator as a function coherency between the at least first oscillator and the oscillation of the at least second oscillator.

Abstract: A method identifies electric load types of a plurality of different electric loads. The method includes providing a load feature database of a plurality of different electric load types, each of the different electric load types including a first load feature vector having at least four different load features; sensing a voltage signal and a current signal for each of the different electric loads; determining a second load feature vector comprising at least four different load features from the sensed voltage signal and the sensed current signal for a corresponding one of the different electric loads; and identifying by a processor one of the different electric load types by determining a minimum distance of the second load feature vector to the first load feature vector of the different electric load types of the load feature database.

Abstract: Methods, systems, and articles of manufacture for annotating of an image are disclosed. These include scoring the image using a plurality of trained classifiers, wherein each of the trained classifiers corresponds to at least one of a plurality of image groups clustered based upon image similarity, and wherein each image group is associated with a set of weighted labels; selecting one or more of the image groups based upon the scoring; aggregating one or more sets of weighted labels associated with the selected one or more image groups; and annotating the image using the aggregated one or more sets of weighted labels.

Abstract: A software optimization system isolates an effect of a change in a control variable from effects of ongoing, unknown changes in other variables. The system discards effects due to noise so that effects of interest to a programmer are more easily visible. The software optimization system treats variations in one or more control variables and in the output of the system as signals. The system varies the control variable at a specific frequency unlikely to correlate with uncontrolled variations in external variables. The system uses digital signal processing (DSP) techniques to filter the output, isolating the frequency of the control variable variation. The system then compares the resulting filtered output to the input to determine the approximate effect of the variation in the control variable.