Abstract: Exemplary systems, apparatus, and methods for evaluating and predicting athletic performance are described. Systems may include a receiver that gathers non-deterministic data on one or more aspects of athletic performance, a deterministic model of the athletic performance, a hybrid processor that creates a conditional probabilistic model from these elements, and a display presenting the evaluated or predicted performance. The system may include sensors affixed to an athlete or their equipment to convey position, acceleration, heart rate, respiration, biomechanical attributes, and detached sensors to record video, audio, and other ambient conditions. Apparatus may include a hybridization processor that communicates the output of conditional probabilistic models directly to athletes, coaches, and trainers using sound, light, or haptic signals, or to spectators using audiovisual enhancements to broadcasts.

Abstract: In implementations of the subject matter described herein, each token for containing an element in the training data is sampled according to a factorization strategy in training. Instead of using a single proposal, the property value of the target element located at the token being scanned is iteratively updated one or more times based on a combination of an element proposal and a context proposal. The element proposal tends to accept a value that is popular for the target element independently of the current piece of data, while the context proposal tends to accept whenever the property value that is popular in the context of the target data or popular for the element itself. The proposed modeling training approach can converge in a quite efficient way.

Abstract: A configuration mapping system and method increase the effectiveness of mapping of information from an established product line to a new product offering. In at least one embodiment, the configuration mapping system herein uses configuration mapping rules to map individual product features and entire configurations from established products to a new product offering. The configuration mapping system also provides a way to appropriately map, for example, demand and sales information for the purpose of demand estimation and sales prediction. Conventionally, mapping can be ineffective because the configuration mapping rules usually focus on one part of the product at a time, and, if applied in isolation, the impact on other parts is missed. The systems and method herein provide a way to integrate configuration mapping rules across feature parts, time periods, and product lines into a unified, holistic view, allowing for new insights.

Abstract: Systems and methods are provided for performing data mining and statistical learning techniques on a big data set. More specifically, systems and methods are provided for linear regression using safe screening techniques. Techniques may include receiving a plurality of time series included in a prediction hierarchy for performing statistical learning to develop an improved prediction hierarchy. It may include pre-processing data associated with each of the plurality of time series, wherein the pre-processing includes tasks performed in parallel using a grid-enabled computing environment. For each time series, the system may determine a classification for the individual time series, a pattern group for the individual time series, and a level of the prediction hierarchy at which the each individual time series comprises an need output amount greater than a threshold amount.

Abstract: A sunset prediction system performs the steps of determining a location of interest, and collecting weather forecast data at or around the time of either sunrise or sunset, at the location and a surrounding area adjacent the location. The collected weather forecast data includes cloud density and cloud altitude. The sun prediction system then determines from the collected weather forecast data whether there will be clouds at the location that are advantageous for photography at or around the time of sunrise or sunset, and whether the sunlight will be obstructed by obstructing clouds in the surrounding area. A forecast rating is then reported so that the user can determine whether to plan photography at the location.

Abstract: A method and system for identifying workplace risk factors is provided. The method includes monitoring via execution of multiple geographically distributed sensor devices, workplace injury based events associated with individuals at a multisite distributed workplace environment. Current injury data describing the workplace injury based events is stored and predicted future workplace injury based events associated with future workplace injury based events with respect to a predicted plurality of individuals at the multisite distributed workplace environment are determined. Injury risk mitigating actions associated with prevention of said predicted future workplace injury based events are generated and an associated cost optimized reduction plan for prioritized implementation of the injury risk mitigating actions is generated.

Abstract: A computer-implemented technique is described herein for modifying original keyword information to increase the probability that it will match the queries input by users. The technique operates by using a search engine to provide supplemental information that is relevant to the original keyword information. The technique then mines the supplemental information to extract frequently-occurring n-grams. Next, the technique removes n-grams that are considered to represent noise, and then uses a deep-structured machine-learned model to assign score values to the remaining n-grams. Finally, the technique supplements and/or replaces the original keyword information with the highest-scoring n-grams.

Abstract: Methods and systems for classifying defects detected on a specimen with an adaptive automatic defect classifier are provided. One method includes creating a defect classifier based on classifications received from a user for different groups of defects in first lot results and a training set of defects that includes all the defects in the first lot results. The first and additional lot results are combined to create cumulative lot results. Defects in the cumulative lot results are classified with the created defect classifier. If any of the defects are classified with a confidence below a threshold, the defect classifier is modified based on a modified training set that includes the low confidence classified defects and classifications for these defects received from a user. The modified defect classifier is then used to classify defects in additional cumulative lot results.

Abstract: Implementations provide for use of spherical random features for polynomial kernels and large-scale learning. An example method includes receiving a polynomial kernel, approximating the polynomial kernel by generating a nonlinear randomized feature map, and storing the nonlinear feature map. Generating the nonlinear randomized feature map includes determining optimal coefficient values and standard deviation values for the polynomial kernel, determining an optimal probability distribution of vector values for the polynomial kernel based on a sum of Gaussian kernels that use the optimal coefficient values, selecting a sample of the vectors, and determining the nonlinear randomized feature map using the sampled vectors.

Abstract: Resource allocation for machine learning is described such as for selecting between many possible options, for example, as part of an efficient training process for random decision tree training, for selecting which of many families of models best describes data, for selecting which of many features best classifies items. In various examples samples of information about uncertain options are used to score the options. In various examples, confidence intervals are calculated for the scores and used to select one or more of the options. In examples, the scores of the options may be bounded difference statistics which change little as any sample is omitted from the calculation of the score. In an example, random decision tree training is made more efficient while retaining accuracy for applications not limited to human body pose detection from depth images.

Abstract: Certain example embodiments relate to techniques for detecting anomalies in streaming data. More particularly, certain example embodiments use an approach that combines both unsupervised and supervised machine learning techniques to create a shared anomaly detection model in connection with a modified k-means clustering algorithm and advantageously also enables concept drift to be taken into account. The number of clusters k need not be known in advance, and it may vary over time. Models are continually trainable as a result of the dynamic reception of data over an unknown and potentially indefinite time period, and clusters can be built incrementally and in connection with an updatable distance threshold that indicates when a new cluster is to be created. Distance thresholds also are dynamic and adjustable over time.

Abstract: A computer system includes a memory storing a data structure representing a neural network. The data structure includes a plurality of fields including values representing topology of the neural network. The computer system also includes one or more processors configured to perform neural network classification by operations including generating a vector representing at least a portion of the neural network based on the data structure. The operations also include providing the vector as input to a trained classifier to generate a classification result associated with at least the portion of the neural network, where the classification result is indicative of expected performance or reliability of the neural network. The operations also include generating an output indicative of the classification result.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating output sequences from input sequences. One of the methods includes obtaining an input sequence having a first number of inputs arranged according to an input order; processing each input in the input sequence using an encoder recurrent neural network to generate a respective encoder hidden state for each input in the input sequence; and generating an output sequence having a second number of outputs arranged according to an output order, each output in the output sequence being selected from the inputs in the input sequence, comprising, for each position in the output order: generating a softmax output for the position using the encoder hidden states that is a pointer into the input sequence; and selecting an input from the input sequence as the output at the position using the softmax output.

Abstract: A system and method for data gathering system, comprising a data-aware knowledge base storing knowledge on relative costs of obtaining various data items; and a data retrieval decision-making processor operative, when an individual data element is sought to be retrieved, to determine whether or not to retrieve the data element by comparing at least one parameter representing need for the data element, also termed herein a utility value, with at least one parameter, retrieved from the data-aware knowledge base, which represents relative cost of obtaining the data element.

Abstract: A method, system, and computer program product for generating database cluster health alerts using machine learning. A first database cluster known to be operating normally is measured and modeled using machine learning techniques. A second database cluster is measured and compared to the learned model. More specifically, the method collects a first set of empirically-measured variables of a first database cluster, and using the first set of empirically-measured variables a mathematical behavior predictor model is generated. Then, after collecting a second set of empirically-measured variables of a second database cluster over a plurality of second time periods, the mathematical behavior predictor model classifies the observed behavior. The classified behavior might be deemed to be normal behavior, or some form of abnormal behavior. The method forms and report alerts when the classification deemed to be anomalous behavior, or fault behavior.

Abstract: A computing system provides distributed training of a neural network model. Explore phase options, exploit phase options, a subset of a training dataset, and a validation dataset are distributed to a plurality of computing devices. (a) Execution of the model by the computing devices is requested using the subset stored at each computing device. (b) A first result of the execution is received from a computing device. (c) Next configuration data for the neural network model is selected based on the first result and distributed to the computing device. (a) to (c) is repeated until an exploration phase is complete. (d) Execution of the neural network model is requested. (e) A second result is received. (f) Next configuration data is computed based on the second result and distributed to the computing device. (d) to (f) is repeated until an exploitation phase is complete. The next configuration data defines the model.

Abstract: Personality aggregation and web browsing is described, for example, to find personality profiles of website audiences for use in recommendation systems, advertising systems, or other web services. In an embodiment natural browsing sequences of users who have given their consent are submitted to a pattern matching process to identify personality trait scores serendipitously occurring in the sequences. In an embodiment, an aggregator combines the personality trait scores by website to obtain audience personality profiles. In an example, a machine learning process carries out the aggregation and enables audience personality profiles of other websites to be predicted. For example, a random decision forest is trained using the natural browsing sequences having identified personality trait scores and once trained, is used to predict personality trait scores of other websites.

Abstract: Methods, systems, and computer program products for correcting existing predictive model outputs with social media features over multiple time scales are provided herein. A method includes examining multiple items of user content derived from one or more social media sources during a given time period to identify one or more items of user content pertaining to a target entity; analyzing the items of user content pertaining to the target entity to determine one or more predictive features in the items of user content related to a target variable associated with the target entity; computing a correlation measure between each of the predictive features and the target variable; and modifying an output of an existing predictive model associated with the target variable via incorporation of the predictive feature with the highest correlation measure into the output of the existing predictive model to generate an updated predictive output.

Abstract: A mechanism is provided in a data processing system for automatically generating question and answer pairs for training a question answering system for a given domain. The mechanism identifies a set of patterns of components in passages within a corpus of documents for the given domain. The mechanism identifies a set of rules that correspond to the set of patterns for generating question and answer pairs from the passages within the corpus of documents. The mechanism applies the set of rules to the passages to generate the question and answer pairs.

Abstract: Embodiments of the present invention provide a method for detecting a temporal change of name associated with performance data. The method comprises receiving at least one candidate name replacement pair comprising a pair of names. The method further comprises, in a training stage, for each known name replacement pair included in the performance data, determining a window of time covering a most recent appearance of a first name of the known name replacement pair. The window of time is determined based on quantitative features of a time series model comprising performance data for the first name and a second name of the known name replacement pair. The method further comprises, in the training stage, training a machine learning classifier based on quantitative features computed using a portion of the performance data for the first name and the second name, where the portion is within the window of time determined.