Abstract: In general, embodiments of the present invention provide systems, methods and computer readable media for an adaptive oracle-trained learning framework for automatically building and maintaining models that are developed using machine learning algorithms. In embodiments, the framework leverages at least one oracle (e.g., a crowd) for automatic generation of high-quality training data to use in deriving a model. Once a model is trained, the framework monitors the performance of the model and, in embodiments, leverages active learning and the oracle to generate feedback about the changing data for modifying training data sets while maintaining data quality to enable incremental adaptation of the model.

Abstract: A holistic hospital patient care and management system comprises a data store operable to receive and store patient data including clinical and non-clinical data; a plurality of video cameras to capture images of the patients; a plurality of presence detection sensors to detect the presence and location of the patients; a risk logic module configured to apply at least one predictive model to the clinical and non-clinical data, including the captured images, to determine at least one risk score associated with the patients; a facial biological change logic module configured to receive location data from the plurality of presence detection sensors, the risk score and medical condition associated with the patients, and captured images of the patients, and generating an alert in response to a detected change in biological change of a patient.

Abstract: A method and apparatus may include receiving a signal from a motor. The signal is received while the motor is operating. The method also includes performing a pre-processing of the signal. The method also includes inputting the signal to a 1D convolutional neural network. The method also includes detecting a fault of the motor based on the output of the neural network.

Abstract: The invention provides, in some aspects, a computer-implemented method for enabling enhanced functionality in a software application. The method includes executing, on a computer, an enhancement engine that is communicatively coupled to a rules base (or other store that contains rules) and/or a rules engine that executes rules (e.g., from the rules base). The enhancement engine receives a request to enable enhanced functionality in an application that is defined, at least in part, by a plurality of such rules, where the request specifies a selected rule in the application for such enhancement. The enhancement engine identifies (or ascertains) a new rule at least partially providing the enhanced functionality and (i) updates the rules base (or other store) to include the new rule along with the others that define at least a portion of the application and/or (ii) effects execution by the rules engine of the new rule along with those others.

Abstract: A novel entity resolution approach for the organization entity domain can be implemented in the MapReduce framework with low memory requirements so that it may scale to large scale datasets. A new clustering approach, sClust, significantly improves the recall of the pairwise classifier.

Abstract: A personalized recommendation model scores each object in an interaction set of objects with which a user interacted and in a ransom set of objects with which the user lacks known interaction. A system sorts each scored object based on a decreasing order of each corresponding score, and identifies a high scoring set of the sorted objects with a number (equal to the number of objects in the interaction set of objects) of highest corresponding scores. The system aggregates a corresponding order value for each object in the high scoring set that is also in the interaction set of objects (the corresponding order value for an object is based on a corresponding order for the object in the high scoring set). The system evaluates the model for the user by dividing the aggregated order value by an aggregation of a corresponding order value for each object in the high scoring set.

Abstract: A system, method and computer program product for automatic document classification, including an extraction module configured to extract structural, syntactical and/or semantic information from a document and normalize the extracted information; a machine learning module configured to generate a model representation for automatic document classification based on feature vectors built from the normalized and extracted semantic information for supervised and/or unsupervised clustering or machine learning; and a classification module configured to select a non-classified document from a document collection, and via the extraction module extract normalized structural, syntactical and/or semantic information from the selected document, and generate via the machine learning module a model representation of the selected document based on feature vectors, and match the model representation of the selected document against the machine learning model representation to generate a document category, and/or classificatio

Abstract: A method of collaborative filtering in combination with time factor includes: establishing an exponential smoothing model; acquiring a time period proposed for the exponential smoothing model, the time period includes a plurality of time cycles; acquiring a plurality of user identifiers and user preference degree values of the user identifiers over a specified product during the plurality of time cycles; performing iterative calculations of the user preference degree values utilizing the exponential smoothing model, and obtaining smoothing results corresponding to the time cycles; generating a sparse matrix utilizing the user identifiers and the smoothing result corresponding to the time cycles, the sparse matrix includes a plurality of user preference degrees to be predicted; acquiring a collaborative filtering model and inputting the smoothing results corresponding to the time cycles into the collaborative filtering model; and training through the collaborative filtering model, calculating and obtaining pre

Abstract: An aspect of the present disclosure aims to reduce problems associated with data acquisition of a rule set. Systems and methods enabling a semantic reasoner to stage acquisition of data objects necessary to bring each of the rules stored in the knowledge base to a conclusion are disclosed. To that end, a dependency chain is constructed, identifying whether and how each rule depends on other rules. Based on the dependency chain, the rules are assigned to difference epochs and reasoning engine is configured to perform machine reasoning over rules of each epoch sequentially. Moreover, when processing rules of each epoch, data objects referenced by the rules assigned to a currently processed epoch are acquired according to a certain order established based on criteria such as e.g. cost of acquisition of data objects. Such an approach provides automatic determination and just-in-time acquisition of data objects required for semantic reasoning.

Abstract: An example method executed by a semantic reasoner is disclosed. The method includes identifying, from a plurality of rules, one or more pairs of chained rules, and, from the one or more pairs of chained rules, assigning rules chained together to a respective rule-set of P rule-sets. The method also includes assigning individuals, from a plurality of individuals referenced by the plurality of rules, referenced by each rule-set of the P rule-sets to an individual-set associated with the each rule-set and mapping the rules from the each rule-set and the individuals from the individual-set associated with the each rule-set into a respective knowledge base instance associated with the each rule-set. Such a method ensures knowledge completeness and sound inference while allowing parallel semantic reasoning within a given stream window.

Abstract: Systems and methods are provided for reducing failure rates of a manufactured products. Manufactured products may be clustered together according to similarities in their production data. Manufactured product clusters may be analyzed to determine mechanisms for failure rate reduction, including adjustments to test quality parameters, product formulas, and product processes. Recommended product adjustments may be provided.

Abstract: Data processing apparatus operative to generate a classification component is disclosed. The data processing apparatus is configured to provide a template classifier bank comprising a plurality of classifier modules, each classifier module operative to receive training data comprising data elements having one of two or more known class affiliations and to output a class affiliation estimate value for each input data element. The data processing apparatus is further configured to derive a combination of the class affiliation estimate values providing a highest correlation to the two or more known class affiliations, and to generate a classification component formed of a resultant classifier bank comprising a combination of the plurality of classifier modules corresponding to the combination of estimate values providing the highest correlation.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for classifying user behavior as anomalous. One of the methods includes obtaining user behavior data representing behavior of a user in a subject system. An initial model is generated from training data, the initial model having first characteristic features of the training data. A resampling model is generated from the training data and from multiple instances of the first representation for a test time period. A difference between the initial model and the resampling model is computed. The user behavior in the test time period is classified as anomalous based on the difference between the initial model and the resampling model.

Abstract: Conversation user interfaces that are configured for virtual assistant interaction may include tasks to be completed that may have repetitious entry of the same or similar information. User preferences may be learned by the system and may be confirmed by the user prior to the learned preference being implemented. Learned preferences may be identified in near real-time on large collections of data for a large population of users. Further, the learned preferences may be based at least in part on previous conversations and actions between the system and the user as well as user-defined occurrence thresholds.

Abstract: Aspects of the disclosure generally relate to computer generated environments, and may be generally directed to devices, systems, methods, and/or applications for learning an avatar's or an application's operating while being at least partially operated by a user and causing an avatar or an application to operate autonomously resembling the user's consciousness or methodology of avatar or application operating. Aspects of the disclosure also generally relate to other disclosed embodiments.

Abstract: Mechanisms are provided for detecting interesting decision rules from a set of decision rules in a tree ensemble. Each tree in the tree ensemble is traversed in order to assign each individual data record from a set of data records to an identified leaf node in each tree. Predicted values are determined for the tree ensemble based on predictions provided by each leaf node to which each individual data record is assigned. Interesting sub-indices for decision rules from the set of decision rules are determined and, for each decision rule corresponding to the leaf nodes in the tree ensemble, the sub-indices are combined into interestingness index It. The decision rules are ranked corresponding to the leaf nodes in the tree ensemble according to the associated value of the interestingness index It and a subset of the decision rules corresponding to the leaf nodes in the tree ensemble are reported.

Abstract: A method for providing cognitive insight via a cognitive information processing system comprising: encapsulating an operation for providing a desired cognitive insight; and, applying the operation to a target cognitive graph to generate a cognitive insight.

Abstract: Mechanisms are provided for detecting interesting decision rules from a set of decision rules in a tree ensemble. Each tree in the tree ensemble is traversed in order to assign each individual data record from a set of data records to an identified leaf node in each tree. Predicted values are determined for the tree ensemble based on predictions provided by each leaf node to which each individual data record is assigned. Interesting sub-indices for decision rules from the set of decision rules are determined and, for each decision rule corresponding to the leaf nodes in the tree ensemble, the sub-indices are combined into interestingness index It. The decision rules are ranked corresponding to the leaf nodes in the tree ensemble according to the associated value of the interestingness index It and a subset of the decision rules corresponding to the leaf nodes in the tree ensemble are reported.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for organizing trained and untrained neural networks. In one aspect, a neural network device includes a collection of node assemblies interconnected by between-assembly links, each node assembly itself comprising a network of nodes interconnected by a plurality of within-assembly links, wherein each of the between-assembly links and the within-assembly links have an associated weight, each weight embodying a strength of connection between the nodes joined by the associated link, the nodes within each assembly being more likely to be connected to other nodes within that assembly than to be connected to nodes within others of the node assemblies.

Abstract: Hierarchical branching deep convolutional neural networks (HD-CNNs) improve existing convolutional neural network (CNN) technology. In a HD-CNN, classes that can be easily distinguished are classified in a higher layer coarse category CNN, while the most difficult classifications are done on lower layer fine category CNNs. Multinomial logistic loss and a novel temporal sparsity penalty may be used in HD-CNN training. The use of multinomial logistic loss and a temporal sparsity penalty causes each branching component to deal with distinct subsets of categories.