Abstract: Facilitate the procedure of evaluating a predictor. This evaluation system comprises an input receiving unit via which elements constituting an evaluation index are specified and an evaluation-index calculation unit that calculates an evaluation-index value for a data set. The evaluation index comprises an element of a first type that evaluates the sample data, an element of a second type that applies weights to the sample data, and an element of a third type that performs a statistical process on a plurality of sample data based on information outputted by the element of the first type and the element of the second type. The evaluation-index calculation unit calculates the above-mentioned evaluation-index value based on the evaluation index comprising the elements received by the input receiving unit.

Abstract: An approach is provided in which a knowledge manager creates a pattern set based on training data that includes paraphrases and a set of first syntactic patterns. The knowledge manager receives a user question and matches one of the first syntactic patterns to a second syntactic pattern generated from the user question. Based on the matching, the knowledge manager generates new questions using the paraphrases in the pattern set and utilizes the new questions to query a second set of data and generate candidate answers that correspond to the user question.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network. One of the methods includes obtaining partitioned training data for the neural network, wherein the partitioned training data comprises a plurality of training items each of which is assigned to a respective one of a plurality of partitions, wherein each partition is associated with a respective difficulty level; and training the neural network on each of the partitions in a sequence from a partition associated with an easiest difficulty level to a partition associated with a hardest difficulty level, wherein, for each of the partitions, training the neural network comprises: training the neural network on a sequence of training items that includes training items selected from the training items in the partition interspersed with training items selected from the training items in all of the partitions.

Abstract: An approach is provided in which a knowledge manager creates a pattern set that includes paraphrases and corresponding paraphrase scores. The paraphrase scores are based on a set of first candidate answers obtained from querying a first set of resource data. The knowledge manager performs a search, which is based on the paraphrases and a user question, on a second set of resource data and identifies a set of second candidate answers. In turn, the knowledge manager scores the set of second candidate answers based on the paraphrase scores corresponding to the paraphrases utilized to identify the set of second candidate answers.

Abstract: An apparatus and method are described for a neuromorphic processor design in which neuron timing information is duplicated on a neuromorphic core.

Abstract: A classification method includes receiving a collection of samples, each sample comprising a multidimensional feature representation. A class label prediction for each sample in the collection is generated with one or more pretrained classifiers. For at least one iteration, each multidimensional feature representation is augmented with a respective class label prediction to form an augmented representation, a set of corrupted samples is generated from the augmented representations, and a transformation that minimizes a reconstruction error for the set of corrupted samples is learned. An adapted class label prediction for at least one of the samples in the collection is generated using the learned transformation and information is output, based on the adapted class label prediction. The method is useful in predicting labels for target samples where there is no access to source domain samples that are used to train the classifier and no access to target domain training data.

Abstract: The embodiments set forth techniques for implementing various “prediction engines” that can be configured to provide different kinds of predictions within a mobile computing device. According to some embodiments, each prediction engine can assign itself as an “expert” on one or more “prediction categories” within the mobile computing device. When a software application issues a request for a prediction for a particular category, and two or more prediction engines respond with their respective prediction(s), a “prediction center” can be configured to receive and process the predictions prior to responding to the request. Processing the predictions can involve removing duplicate information that exists across the predictions, sorting the predictions in accordance with confidence levels advertised by the prediction engines, and the like. In this manner, the prediction center can distill multiple predictions down into an optimized prediction and provide the optimized prediction to the software application.

Abstract: Human Computer Interfaces (HCI) may allow a user to interact with a computer via a variety of mechanisms, such as hand, head, and body gestures. Various of the disclosed embodiments allow information captured from a depth camera on an HCI system to be used to recognize such gestures. Particularly, the HCI system's depth sensor may capture depth frames of the user's movements over time. To discern gestures from these movements, the system may group portions of the user's anatomy represented by the depth data into classes. “Features” which reflect distinguishing features of the user's anatomy may be used to accomplish this classification. Some embodiments provide improved systems and methods for generating and/or selecting these features. Features prepared by various of the disclosed embodiments may be less susceptible to overfitting training data and may more quickly distinguish portions of the user's anatomy.

Abstract: The present invention relates to a method for incremental learning of a classification model, where pre-defined weak incremental learners are distributed over the distinct regions in a set of partitionings of the input domain. The partitionings and regions are organized via a binary tree and they are allowed to vary in a data-driven way, i.e., in a way to minimize the classification error rate. Moreover, to test a given data point, a mixture of decisions is obtained through the models learned in the regions that this point falls in. Hence, naturally, in the cold start phase of the data stream, the simpler models belonging to the larger regions are favored and as more data get available, the invention automatically puts more weights on the more complex models.