Abstract: Systems, methods, and devices are described for implementing learning algorithms on data sets. A data set may be partitioned into a plurality of data partitions that may be distributed to two or more processors, such as a graphics processing unit. The data partitions may be processed in parallel by each of the processors to determine local counts associated with the data partitions. The local counts may then be aggregated to form a global count that reflects the local counts for the data set. The partitioning may be performed by a data partition algorithm and the processing and the aggregating may be performed by a parallel collapsed Gibbs sampling (CGS) algorithm and/or a parallel collapsed variational Bayesian (CVB) algorithm. In addition, the CGS and/or the CVB algorithms may be associated with the data partition algorithm and may be parallelized to train a latent Dirichlet allocation model.

Abstract: Embodiments of systems and methods can determine evaluations of the quality of task results provided by judges. Certain embodiments can determine the quality evaluations with relatively little overlap of judges (e.g., two or three judges). The quality evaluations may include determining an error rate or a confidence index for a judge or for a particular judgment made by the judge. Certain embodiments may determine the quality evaluations by intercomparing judging results from multiple judges without requiring human intervention, comparison to evaluation data known (or assumed) to be accurate, or input from an external judge review panel. In some implementations, task results can be combined with a confidence score (based at least in part on the quality evaluation of the judge who provided the task result), and this information can be used as training data to improve performance of machine learning algorithms.

Abstract: Described herein are methods for training a machine learning-based predictor of affective response to stimuli. The methods involve receiving samples comprising temporal windows of token instances to which a user was exposed, and target values representing affective response annotations of the user in response to the temporal windows of token instances. This data is used for the training of the predictor along with values indicative of the number of the token instances in the temporal windows of token instances, which are used to compensate for non-linear effects resulting from saturation of the user.

Abstract: Methods, systems and media are taught utilizing ranking techniques in machine learning to learn a ranking function. Specifically, ranking algorithms are applied to learn a ranking function that advantageously minimizes ranking error as a function of targeted ranking order discrepancies between a predetermined first ranking of a training plurality of data elements and a second ranking of the training plurality of data elements by the ranking function. The ranking algorithms taught may be applied to ranking representations of chemical structures and may be particularly advantageous in the field of drug discovery, e.g., for prioritizing chemical structures for drug screenings.

Abstract: Exemplary embodiments provide systems, devices, one or more non-transitory computer-readable media and computer-executable methods for managing publication of online advertising. In exemplary embodiments, computer-based publication techniques may include, but is not limited to, automatically determining whether the content of a particular web page article is suitable or unsuitable for accompaniment with one or more advertisements, automatically determining whether an advertisement is suitable or unsuitable for publication on a web page associated with a web page article, and automatically determining a category that may be used to classify the content of a web page article in order to select one or more categories of advertisements suitable for accompaniment with the web page article.

Abstract: Multiple data prediction strategies are received. Each data prediction strategy may predict a next data value in a sequence of data values with a corresponding confidence value. Rather than rely on a single prediction strategy, the predictions of each of the data prediction strategies are linearly combined to generate a single prediction that is more accurate and has a lower overall loss than any of the individual prediction strategies. Further, a deviation is calculated based on the values in the sequence of values that have been observed so far using a weighted sum that favors more recent values in the sequence over less recent values in the sequence. A prediction of the next value in the sequence is generated based on the combined strategies and the calculated deviation.

Abstract: A method and apparatus store media data on a portable electronic device. The method can include monitoring media data storage activity regarding user media data storage activity usage patterns that store media data in a memory in a portable electronic device. The method can include storing the media data storage activity data in the portable electronic device. The method can include triggering a full memory prediction algorithm based on a user action event that affects media data storage in the memory and can include running the full memory prediction algorithm in response to the trigger. The full memory prediction algorithm can predict when the memory will be substantially full based on past user media data storage behavior according to the media data storage activity data. The method can include outputting a potential upcoming full memory indication.

Abstract: Technologies pertaining to learning a computer-executable imitation system that imitates behavior of an existing computer-executable system are described herein. Behavior of an existing computer-executable system can be monitored through monitoring data input to the existing computer-executable system and data output by the existing computer-executable system responsive to receipt of the input data. An imitation system that imitates the behavior of the existing system can be learned, wherein the imitation system comprises a relational model.

Abstract: A computing device receives a document that was incorrectly classified as sensitive data based on a machine learning-based detection (MLD) profile. The computing device modifies a training data set that was used to generate the MLD profile by adding the document to the training data set as a negative example of sensitive data to generate a modified training data set. The computing device then analyzes the modified training data set using machine learning to generate an updated MLD profile.

Abstract: In the area of storage management, service automation can be realized through the use of “MAPE” loop(s). A Planner (P) interacts with the Monitoring (M), Analysis (A) and Execution (E) components in a closed loop. For each new option or potential planning action the Planner (P) invokes the Analysis (A) component. The correctness, as well as effectiveness, of the planning decision is dependent on the Analysis (A) component. Embodiments can utilize an adaptive Analysis (A) component (i.e., an analysis component that can be retrained) that also associates a value of confidence and a corresponding error in the evaluation along with a predicted impact. The Planner (P) component uses this additional information for quoting the final impact of a particular planning action as part of an adaptive MAPE loop to provide improved resource utilization and resource management.

Abstract: Techniques for creating a distributed application flow from a set of rules are provided. The techniques include creating a control-flow graph for each rule, creating one or more dependency links between two or more rules, partitioning a resulting graph, wherein the resulting graph comprises one or more control-flow graphs and one or more dependency links, into one or more operators by determining an optimal set of one or more cuts through the resulting graph such that a cost function is minimized, and generating stream processing flow code from the partitioned graph.

Abstract: An apparatus for classifying an audio signal configured to: estimate at least one shaping parameter value for a plurality of samples of the audio signal; generate at least one audio signal classification value by mapping the at least one shaping parameter value to one of at least two interval estimates; and determine at least one audio signal classification decision based on the at least one audio signal classification value.

Abstract: An apparatus, system, and method are disclosed for defining normal usage of a computing system resource. A method for defining normal usage of a computing system resource includes receiving a repeating schedule that represents system usage of one or more computing resources and receiving one or more demand events that will affect the system usage of the one or more computer resources. The method also automatically creates a predictive temporal profile that represents the system usage of the one or more computer resources from information comprising the repeating schedule and the one or more demand events. The predictive temporal profile is displayed for the user to review.

Abstract: A personalized page rank computation system is described herein that provides a fast MapReduce method for Monte Carlo approximation of personalized PageRank vectors of all the nodes in a graph. The method presented is both faster and less computationally intensive than existing methods, allowing a broader scope of problems to be solved by existing computing hardware. The system adopts the Monte Carlo approach and provides a method to compute single random walks of a given length for all nodes in a graph that it is superior in terms of the number of map-reduce iterations among a broad class of methods. The resulting solution reduces the I/O cost and outperforms the state-of-the-art FPPR approximation methods, in terms of efficiency and approximation error. Thus, the system can very efficiently perform single random walks of a given length starting at each node in the graph and can very efficiently approximate all the personalized PageRank vectors.

Abstract: A novel domain adaption/transfer learning method applied to the problem of classifying abbreviated documents, e.g., short text messages, instant messages, tweets. The method uses a large number of multi-labeled examples (source domain) to improve the learning on the partial observations (target domain). Specifically, a hidden, higher-level abstraction space is learned that is meaningful for the multi-labeled examples in the source domain. This is done by simultaneously minimizing the document reconstruction error and the error in a classification model learned in the hidden space using known labels from the source domain. The partial observations in the target space are then mapped to the same hidden space, and classified into the label space determined by the source domain.

Abstract: A computing device receives a rule that includes information describing conditions associated with a consequence, and identifies rule components corresponding to the rule. The computing device creates a rule formula, based on the rule components, by creating a first-order logic version of the rule and creating a rule formula table based on the first-order logic version of the rule. The computing device stores the rule formula table in a relational database.

Abstract: Methods and systems for diagnosing and identifying a treatment for an orthodontic condition can include a server configured to receive patient data through a website. Methods and systems can include the use of a database that includes or has access to information derived from textbooks and scientific literature and dynamic results derived from ongoing and completed patient treatments. Methods and systems can include the operation of at least one computer program within the server, which can be capable of analyzing patient data and identifying at least one diagnosis of an orthodontic condition. Methods and systems can include assigning a probability value to at least one diagnosis, and the probability value can represent a likelihood that a diagnosis is accurate. Methods and systems can include instructing a computer program to identify at least one treatment approach, a corrective appliance, or a combination thereof for the at least one diagnosis.

Abstract: A method for adjusting a user's experience of a controllable event including determining a user somatic state, using a computer device, from user sensor data collected from at least one physiological sensor; determining a user cognitive state, using the computer device, from user experience data collected from the user; determining a user experience model, using the computer device, from the user somatic state and the user cognitive state; correlating, using the computer device, at least one user hypothesis with the user experience model; and adjusting the controllable event, using the computer device, based upon the at least one user hypothesis.

Abstract: A novel domain adaption/transfer learning method applied to the problem of classifying abbreviated documents, e.g., short text messages, instant messages, tweets. The method uses a large number of multi-labeled examples (source domain) to improve the learning on the partial observations (target domain). Specifically, a hidden, higher-level abstraction space is learned that is meaningful for the multi-labeled examples in the source domain. This is done by simultaneously minimizing the document reconstruction error and the error in a classification model learned in the hidden space using known labels from the source domain. The partial observations in the target space are then mapped to the same hidden space, and classified into the label space determined by the source domain.

Abstract: A system, method, and computer program product are provided for updating an algorithm. In use, feedback data associated with an identification of an object is received. Additionally, one or more algorithms are updated, based on the feedback data. Further, the updated algorithms are distributed.