Abstract: A computer-implement method and an apparatus are provided for neural network reinforcement learning. The method includes obtaining, by a processor, an action and observation sequence. The method further includes inputting, by the processor, each of a plurality of time frames of the action and observation sequence sequentially into a plurality of input nodes of a neural network. The method also includes updating, by the processor, a plurality of parameters of the neural network by using the neural network to approximate an action-value function of the action and observation sequence.

Abstract: A facility for identifying combinations of feature and machine learning algorithm parameters, where each combination can be combined with one or more machine learning algorithms to train a model, is disclosed. The facility evaluates each genome based on the ability of a model trained using that genome and a machine learning algorithm to produce accurate results when applied to a validation data set by, for example, generating a fitness or validation score for the trained model and the corresponding genome used to train the model. Genomes that produce fitness scores that exceed a fitness threshold are selected for mutation, mutated, and the process is repeated. These trained models can then be applied to new data to generate predictions for the underlying subject matter.

Abstract: A neural network device includes a crossbar grid including first metal lines running in a first direction and second metal lines running transversely to the first metal lines and being electrically isolated from the first metal lines. An array of cross-over elements is included. Each cross-over element is connected between a first metal line and a second metal line. The cross-over elements each include a floating gate transistor device having a floating node. The floating node is configured to store a programmable weight value.

Abstract: A big data processing method based on a deep learning model satisfying K-degree sparse constraints comprises: step 1), constructing a deep learning model satisfying K-degree sparse constraints using an un-marked training sample via a gradient pruning method, wherein the K-degree sparse constraints comprise a node K-degree sparse constraint and a level K-degree sparse constraint; step 2), inputting an updated training sample into the deep learning model satisfying the K-degree sparse constraints, and optimizing a weight parameter of each layer of the model, so as to obtain an optimized deep learning model satisfying the K-degree sparse constraint; and step 3), inputting big data to be processed into the optimized deep learning model satisfying the K-degree sparse constraints for processing, and finally outputting a processing result. The method in the present invention can reduce the difficulty of big data processing and increase the speed of big data processing.

Abstract: A crime prediction server accesses an SNS server, collects submission information, which includes crime-related terms, as crime-related information from the submission information of general citizens, calculates statistical data for each attribute, which includes an occurrence place, a crime occurrence hour, a crime type of a crime, for the crime-related information, and sends crime data and map data, which are extracted from the statistical data of the crime-related information, in response to a request from a center device. The center device superimposes the crime data on the map data for each attribute on a display, and plots and displays the crime data in a position corresponding to the crime occurrence place on a map.

Abstract: An information processor is provided that includes an inference module configured to extract a subset of data from information in an input and a classification module configured to classify the information in the input based on the extracted data. The inference module includes a first plurality of convolvers acting in parallel to apply each of N1 convolution kernels to each of N2 portions of the input image in order to generate an interim sparse representation of the input and a second plurality of convolvers acting in parallel to apply each of N3 convolution kernels to each of N4 portions of the interim sparse representation to generate a final sparse representation containing the extracted data. In order to take advantage of sparsity in the interim sparse representation, N3 is greater than N4 to parallelize processing in a non-sparse dimension and/or the second plurality of convolvers comprise sparse convolvers.

Abstract: Embodiments of the present invention provide a system for predicting one or more events and mitigating the impact of the one or more events. The system is typically configured for presenting a list of exposures to a user via an event prediction application user interface, prompting the user to select an exposure from the list of exposures, receiving a selection of the exposure and at least one option from the user device, determining type of the exposure based on the at least one option, determining one or more distribution models based on the type of the exposure, estimating occurrence of the one or more events associated with the exposure using at least one distribution model from the one or more distribution models, and in response to estimating the occurrence of the one or more events, triggering one or more automated counter measures.

Abstract: One embodiment provides for a non-transitory machine readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising providing an interface to define a neural network using machine-learning domain specific terminology, wherein the interface enables selection of a neural network topology and abstracts low-level communication details of distributed training of the neural network.

Abstract: A method includes training a prediction model to forecast a likelihood of curtailment for at least one wind turbine. The prediction model is trained, by a processor system, using historical information and historical instances of curtailment. The method also includes forecasting the likelihood of curtailment for the at least one wind turbine using the trained prediction model. The method also includes outputting the forecasted likelihood.

Abstract: In an approach for providing a response based on situational context, a computer determines that an individual is within a proximity of a computing device. The computer identifies an identity associated with the determined individual within the proximity of the computing device. The computer determines a location associated with the identified identity. The computer identifies an entry within a table based on at least in part on the identified identity and the determined location. The computer determines a stimulus associated with the identified entry occurs. The computer provides a response based on determining the stimulus associated with the identified entry occurs.

Abstract: Described herein are techniques for dealing with the problem of security vulnerabilities in computer software due to undefined behavior that may be exploited by attackers. A machine learning (ML) model is used for detecting an exploit execution within a given trace of application execution. In a specific embodiment, the ML model identifies whether there is any gadget or gadget-chain execution at branch points of a subject program.

Abstract: A system for identifying compatible meal options. The system includes a body analysis module configured to receive a user biological marker, select a clustering dataset from a clustering database, generate a hierarchical clustering algorithm and assign a plurality of user body measurements to a first classified dataset cluster. The system includes a food analysis module configured to select a food training set from a food database, generate using a supervised machine-learning process a food model, generate a food tolerance instruction set, and display on a graphical user interface the food tolerance instruction set. The system includes a menu generator module configured to select a menu training set from a menu database, generate using a supervised machine-learning process a menu model that produces an output containing a plurality of menu options, and display on a graphical user interface the plurality of menu options.

Abstract: A voice activated knowledge management system may be used as a virtual assistant. In some cases, a knowledge management system may be configured to receive a voice request from a user, generate and send a knowledge base query to each of the two or more different knowledge base engines, and fuse the resulting responses from the knowledge base engines, resulting in a fused response. The fused response may be provided back to the user as a response to the voice request and/or may be provided as a device command to control a corresponding device.

Abstract: Provided is a self-correcting stock price movement predictor built with Artificial Intelligence (AI) techniques that are for empirical data and represent trends on a particular day for an identified stock. Robots, Internet bots, and so forth, are used to source events in real time in the World Wide Web and which may have a potential impact on the movement of the stock. When the web is spidered (or browsed), a device is able to capture the data, which might have an impact on the stock. The data may be structured and ranked using various techniques, including AI techniques. These techniques are applied in order to predict the stock behavior and the movement of that particular stock. An output of the tool is provided to automatically determine an action to take relative to an identified stock.

Abstract: Some embodiments are associated with a system and method for deep learning unsupervised anomaly detection in Internet of Things (IoT) sensor networks or manufacturing execution systems. The system and method use an ensemble of a plurality of generative adversarial networks for anomaly detection.

Abstract: A method of machine learning that teaches a computer to determine likelihood that a medical journal article is classified as high value for an intended system. In some embodiments, the method includes procuring, in a medical articles database on the computer, a training set including medical articles recommended by subject matter experts associated with medical journals that have published the medical articles. The method can also include identifying, by a feature extraction controller of the computer, first features in the medical articles via a remote annotation service and a remote article information service. The method can also include identifying, by intended system services electronically available to the intended system, second features in the medical articles. The method can also include reducing, by the feature extraction controller, the first and second features to form a set of relevant features.

Abstract: Example embodiments provide a system and method for providing optimized channel change using probabilistic modeling. A digital receiver detects an occurrence of a channel event. In response to the detected occurrence, the digital receiver accesses a probabilistic causal model from a data storage device, and dynamically learns in real time one or more probabilities based on the detected channel event and the probabilistic causal model. The digital receiver updates the probabilistic causal model at the data storage device with the learned one or more probabilities. A next channel is determined by the digital receiver based on the updated probabilistic causal model.

Abstract: An apparatus acquires learning-data, including feature-elements, to which a label is assigned. The apparatus generates a first-set of expanded feature-elements by expanding the feature-elements. With reference to a model where a confidence value is stored in association with each of a second-set of expanded feature-elements, the apparatus updates confidence values associated with expanded feature-elements common between the first- and second-sets of expanded feature-elements, based on the label.

Abstract: Systems and methods for time series prediction are described. The systems and methods include encoding driving series into encoded hidden states, the encoding including adaptively prioritizing driving series at each timestamp using input attention, the driving series including data sequences collected from sensors. The systems and methods further includes decoding the encoded hidden states to generate a predicting model, the decoding including adaptively prioritizing encoded hidden states using temporal attention. The systems and methods further include generating predictions of future events using the predicting model based on the data sequences. The systems and methods further include generating signals for initiating an action to devices based on the predictions.

Abstract: An online system receives a request from a third-party application for a content item to be delivered to a user of a user device, the request including a device identifier corresponding to the user device used to access the third-party application. In order to deliver a targeted content item, the online system accesses inference model to infer an identity of the user, based upon device usage parameters based upon the received request, historical data associated with the device identifier, or an indication of activity associated with the device identifier on the online system or one or more first-party applications.