Abstract: Embodiments of the invention provide a method, system, and computer readable medium for classifying network traffic based on application signatures generated during a training phase. The application signatures are generated based on tokens extracted from a training set that is generated by a particular application during the training phase. Accordingly, a new token extracted in real-time from current network data is compared to the application signatures to determine if the current network data is generated by the particular application.

Abstract: Systems and methods are provided for training a model using machine learning. An exemplary method may include providing, by the model in a training session, an action to an environment to receive feedback from the environment. The method may also include generating, by a behavior simulator, a plurality of predicted outcomes from the environment resulting from the action. The method may further include training the model, using at least a subset of the predicted outcomes, to generate a set of candidate models. The method may include receiving actual feedback from the environment and determining whether the actual feedback matches one of the predicted outcomes in the subset. Responsive to the determination that the actual feedback matches one of the predicted outcomes in the subset, the method may include using, in a new training session, the candidate model in the set corresponding to the matched predicted outcome.

Abstract: In a general aspect of the examples described, sensor data from a sensor device are used to improve training data for a motion detection system. In some aspects, “ground truth” labels, sensor data, and channel information are received for a motion detection training period. The labels and sensor data are analyzed to detect variances between the labels and sensor data. The labels are corrected or augmented based on the sensor data. The channel information is tagged with the labels and provided as training information to train the motion detection system.

Abstract: Disclosed is a method of providing a disease co-occurrence probability including (a) receiving a disease network in which respective diseases are shown as nodes and a correlation between diseases is shown as an edge between the nodes and (b) calculating, when at least one disease is given, a probability of an occurrence of another disease in addition to the given disease, the corresponding disease which accompanies the given disease, from the disease network.

Abstract: Methods and arrangements for forecasting at least one attribute of a future event based on a repository of historical event data associated with historical events comparable to the future event. A plurality of event data points are obtained from the repository of historical event data. The plurality of event data points are grouped in accordance with at least one category and a plurality of subcategories to create at least one data tree. Certain of the grouped event data points are designated to form a set of candidate data attributes, and the designated set of candidate data attributes are compared to a set of data attributes from the at least one data tree associated with the future event. Based on the comparing, there is identified a data attribute missing from the set of data attributes from the at least one data tree associated with the future event, and a value for the missing data attribute is forecast. Other variants and embodiments are broadly contemplated herein.

Abstract: A method may include determining, by a computing device and based on at least one user coherency factor, a user coherency level. The coherency level may include a predicted ability of a user to comprehend information. The method may also include determining, by the computing device and based on the user coherency level, information having a complexity that satisfies the predicted ability of the user to comprehend information. The method may further include outputting, by the computing device, at least a portion of the information.

Abstract: A medical diagnosis support apparatus is provided. In the medical diagnosis support apparatus, an acquisition unit acquires medical information associated with a diagnosis target as input information. An inference unit infers a diagnosis name of the diagnosis target based on the acquired input information. A calculation unit calculates the influence rate of each input information with respect to each inference. A creation unit creates a report sentence based on the calculated influence rate.

Abstract: An event-notification system provides for monitoring resource-status parameters so as to repeatedly generate resource-status values for each of N resource-status parameters applicable to the resources. Each datapoint specifies a resource, a time of collection, and the values of parameters for the resource at the time. The datapoints are represented in an N-dimensional space, including N dimensions corresponding to the resource-status parameters and one dimension devoted to time. The datapoints are clustered on a proximity basis. The clusters are manually classified as Normal or Ack-Abnormal. A machine-learning engine develops a model that allows the classifications to be automated. Alerts are sent for respective Abnormal clusters rather than for each abnormal datapoint, thus reducing the amount of data an administrator must deal with to address potential problems.

Abstract: A method for generating data explanations in a recursive cortical network includes receiving a set of evidence data at child feature nodes of a first layer of the recursive cortical network, setting a transformation configuration that directs messaging of evidence data and transformed data between layers of the network, performing a series of transformations on the evidence data according to the transformation configuration, the series including at least one forward transformation and at least one reverse transformation, and outputting the transformed evidence data.

Abstract: In one embodiment, a system includes a data navigation unit configured to navigate through a data structure stored in a first memory to a first representation of at least one rule. The system further includes at least one rule processing unit configured to a) receive the at least one rule based on the first representation of the at least one rule from a second memory to one of the rule processing unit, and b) processing a key using the at least one rule.

Abstract: Embodiments disclosed herein provide a system, method, and computer program product for providing a triage classification system. The triage classification system uses a computer model that is developed using historical patient data. The developed computer model is applied to collected patient attribute data from a patient in a pre-hospital setting to generate a triage category. Based on the generated triage category, health care professionals can take desired actions, such as transporting the patient to a facility matching the generated triage category.

Abstract: A graphical interface module may provide a set of graphical presentations comprising at least: a Likelihood of Delivery chart showing a probability distribution of predicted delivery dates; a Delivery Date Risk Trend chart showing how the completion time for the project predicted according to the Likelihood of Delivery chart has changed over time; and a Burndown chart that shows at least work-items of planned work for the project. Each of the Likelihood of Delivery chart, the Delivery Date Risk Trend chart, and the Burndown chart has a timeline axis.

Abstract: The technology described in this document can be embodied in a computer-implemented method that includes receiving identification information associated with a geographic location. The identification information includes one or more features that affect an acoustic environment of the geographic location at a particular time. The method also includes determining one or more parameters representing at least a subset of the one or more features, and estimating at least one acoustic parameter that represents the acoustic environment of the geographic location at the particular time. The at least one parameter can be estimated using a mapping function that generates the estimate of the at least one acoustic parameter as a weighted combination of the one or more parameters. The method further includes presenting, using a user-interface displayed on a computing device, information representing the at least one acoustic parameter estimated for the geographic location for the particular time.

Abstract: An information providing method of a server, including: acquiring at least one of content information corresponding to content displayed on a display apparatus and user information of a user of the display apparatus; extracting at least one question related to the content from a question-answer database based on the at least one of the content information and the user information; and transmitting the extracted at least one question to the display apparatus, wherein the question-answer database is generated based on a question-answer template.

Abstract: Systems and methods for optimal sizing of one or more grid-scale batteries for frequency regulation service, including determining a desired battery output power for the one or more batteries for a particular period of time. A battery size is optimized for the one or more batteries for the particular period of time, and the optimizing is repeated using different time periods to generate a set of optimal battery sizes based on at least one of generated operational constraints or quality criteria constraints for the one or more batteries. A most optimal battery is selected from the set of optimal battery sizes.

Abstract: The technology disclosed describes systems and methods for generating feature vectors from resource description framework (RDF) graphs. Machine learning tasks frequently operate on vectors of features. Available systems for parsing multiple documents often generate RDF graphs. Once a set of interesting features to be considered has been established, the disclosed technology describes systems and methods for generating feature vectors from the RDF graphs for the documents. In one example setting, a machine learning system can use generated feature vectors to determine how interesting a news article might be, or to learn information-of-interest about a specific subject reported in multiple articles. In another example setting, viable interview candidates for a particular job opening can be identified using feature vectors generated from a resume database, using the disclosed systems and methods for generating feature vectors from RDF graphs.

Abstract: Compressing a machine learning network, such as a neural network, includes replacing one layer in the neural network with compressed layers to produce the compressed network. The compressed network may be fine-tuned by updating weight values in the compressed layer(s).

Abstract: A brain machine interface (BMI) to control a device is provided. The BMI has a neural decoder, which is a neural to kinematic mapping function with neural signals as input to the neural decoder and kinematics to control the device as output of the neural decoder. The neural decoder is based on a continuous-time multiplicative recurrent neural network, which has been trained as a neural to kinematic mapping function. An advantage of the invention is the robustness of the decoder to perturbations in the neural data; its performance degrades less—or not at all in some circumstances—in comparison to the current state decoders. These perturbations make the current use of BMI in a clinical setting extremely challenging. This invention helps to ameliorate this problem. The robustness of the neural decoder does not come at the cost of some performance, in fact an improvement in performance is observed.

Abstract: Behavioral prediction for targeted end users is described. In one or more example embodiments, a computer-readable storage medium has multiple instructions that cause one or more processors to perform multiple operations. Targeted selectstream data is obtained from one or more indications of data object requests corresponding to a targeted end user. A targeted directed graph is constructed based on the targeted selectstream data. A targeted graph feature vector is computed based on one or more invariant features associated with the targeted directed graph. A behavioral prediction is produced for the targeted end user by applying a prediction model to the targeted graph feature vector. In one or more example embodiments, the prediction model is generated based on multiple graph feature vectors respectively corresponding to multiple end users. In one or more example embodiments, a tailored opportunity is determined responsive to the behavioral prediction and issued to the targeted end user.

Abstract: Method and system for assessing the suitability of an entity using a proxy. A description of a behavior associated with a desirable audience is received. A proxy behavior estimated to be characteristic of the desirable audience is selected. The proxy behavior comprises the performance of proxy events related to the consumption of media received by an entity over a network, which can be found in an entity's consumption history. An entity can be assessed for inclusion in a proxy audience, by examining the entity's consumption history for proxy behaviors. A behavioral model is built using a training set comprising the proxy audience. By applying the behavioral model to the consumption history of a specified entity, the specified entity's suitability for selection can be determined. Advantageously, in an embodiment, the invention enables the use of behavioral modeling techniques even when the complete behavior of the desirable audience is not available.