Abstract: A lightened neural network method and apparatus. The neural network apparatus includes a processor configured to generate a neural network with a plurality of layers including plural nodes by applying lightened weighted connections between neighboring nodes in neighboring layers of the neural network to interpret input data applied to the neural network, wherein lightened weighted connections of at least one of the plurality of layers includes weighted connections that have values equal to zero for respective non-zero values whose absolute values are less than an absolute value of a non-zero value. The lightened weighted connections also include weighted connections that have values whose absolute values are no greater than an absolute value of another non-zero value, the lightened weighted connections being lightened weighted connections of trained final weighted connections of a trained neural network whose absolute maximum values are greater than the absolute value of the other non-zero value.

Abstract: Generating insight on a set of data is provided. A request for information regarding a specific topic is received from a client device corresponding to a requester. An analysis is performed on the request and a type of the information requested is determined based on the analysis. A set of information vendors is selected from a plurality of known information vendors based on the type of the information requested and other factors. Insights on the type of the information requested are obtained from the selected set of information vendors and an analysis is performed on the insights. A response to the request is generated based on the analysis of the insights on the type of the information requested that was obtained from the selected set of information vendors. The response to the request is sent to the client device corresponding to the requester.

Abstract: Methods, systems, and apparatus for prioritizing communications are described. Metadata that characterizes an electronic communication is obtained and a machine learning algorithm is applied to the metadata to generate a scoring model. A score for the electronic communication is generated based on the scoring model.

Abstract: At a machine learning service, an indication of a training data set for a model is obtained. One or more training iterations of the model are conducted using an adaptive input sampling strategy. In a particular iteration, index values for a set of training observations are selected based on a set of sampling weights, parameters of the model are updated based on results using training observations identified by the index values, and sampling weights are modified. A result obtained from a trained version of the machine learning model is provided.

Abstract: Various embodiments of the present invention relate generally to systems and processes for transforming a style of video data. In one embodiment, a neural network is used to interpolate native video data received from a camera system on a mobile device in real-time. The interpolation converts the live native video data into a particular style. For example, the style can be associated with a particular artist or a particular theme. The stylized video data can viewed on a display of the mobile device in a manner similar to which native live video data is output to the display. Thus, the stylized video data, which is viewed on the display, is consistent with a current position and orientation of the camera system on the display.

Abstract: A capability to identify and visualize relationships and commonalities amongst record entities is provided. A plurality of entities are extracted from one or more records. Each extracted entity is associated with a respective feature vector within a vector space of a feature matrix. The feature vectors are distributed within the feature matrix based on semantic relationships amongst the entities of a corpus. Multidimensional coordinates within a dimensionally-reduced vector space of the feature matrix are generated for each extracted entity. One or more cells of a cellular presentation of the feature matrix are identified such that each identified cell represents one or more respective extracted entities. Each cell represents (i) a respective range of multidimensional coordinates within the dimensionally-reduced vector space of the feature matrix and (ii) one or more feature vectors of the plurality of feature vectors within the feature matrix.

Abstract: The disclosure below describes a knowledge pattern machine that goes beyond and is distinct from a traditional search engine as simple information aggregator. Rather than acting as a search engine of the data itself, the knowledge pattern machine use variously layers of artificial intelligence to discover correlations within the queries and historical data, and to derive and recognize data patterns based on user queries for predictively generating new knowledge items or reports that are of interest to the user. Previous patterns and knowledge items or reports are accumulated and incorporated in identification of new data patterns and new predictive knowledge items or reports in response to future user queries, thus providing a stateful machine. The predictive knowledge items are updated in real-time without user interference as the underlying data sources evolve overtime.

Abstract: Methods and apparatuses are described for generation of a data model for identifying and classifying training needs of individuals. A computer data store stores unstructured text. A server computing device generates a vector for search queries in the unstructured text, and generates a training course classification data model that comprises a multi-layered neural network. The server computing device executes the training course classification model using the vectors as input to generate a training course recommendation output vector. The server computing device updates the training course classification data model based upon a rating value for a training course.

Abstract: A method and system are provided. The method includes obtaining, by a hardware processor, candidate data representing a plurality of candidates. The method further includes calculating, by the hardware processor, for each of the candidates, a temporal next state of a Recurrent Neural Network (RNN) by inputting a corresponding one of the candidates to the RNN at a current state. The method also includes merging, by the hardware processor, the temporal next state for each of the candidates to obtain the temporal next state of the RNN.

Abstract: A method of semantic searching. The method may include receiving a first search query, obtaining a disambiguation term for semantically disambiguating the first search query, and creating, with a processor, a second search query based at least in part on the first search query and the disambiguation term. The method may also include at least one of outputting search results obtained from a search conducted based at least in part on the second search query and sending the second search query to a search service for outputting search results.

Abstract: A first training participant performs an iterative process until a predetermined condition is satisfied, where the iterative process includes: obtaining, using secret sharing matrix addition and based on the current sub-model of each training participant and a corresponding feature sample subset of each training participant, a current prediction value of the regression model for a feature sample set, where the corresponding feature sample subset of each training participant is obtained by performing vertical segmentation on the feature sample set; determining a prediction difference between the current prediction value and a label corresponding to the current prediction value; sending the prediction difference to each second training participant; and updating a current sub-model of the first training participant based on the current sub-model of the first training participant and a product of a corresponding feature sample subset of the first training participant and the prediction difference.

Abstract: A method for selecting an action, includes reading, into a memory, a Partially Observed Markov Decision Process (POMDP) model, the POMDP model having top-k action IDs for each belief state, the top-k action IDs maximizing expected long-term cumulative rewards in each time-step, and k being an integer of two or more, in the execution-time process of the POMDP model, detecting a situation where an action identified by the best action ID among the top-k action IDs for a current belief state is unable to be selected due to a constraint, and selecting and executing an action identified by the second best action ID among the top-k action IDs for the current belief state in response to a detection of the situation. The top-k action IDs may be top-k alpha vectors, each of the top-k alpha vectors having an associated action, or identifiers of top-k actions associated with alpha vectors.

Abstract: A solution generation and planning system uses Artificial Intelligence (AI) techniques such as machine learning (ML) data models, predictive analytics and natural language processing (NLP) techniques for generating outputs to aid decision making in the domain of public infrastructure development. The problem statement is analyzed using the NLP techniques to generate word tokens which are employed in identifying issues that aid in selection of appropriate data sources from a plurality of discrete data sources. In addition, data models trained to produce probable solutions for the issue are also selected. The probable solutions are presented to the user who selects one of the probable solutions for implementation. Feedback from the implementation is also incorporated so that the data models are updated per the latest information obtained from the implementation of the user-selected solution.

Abstract: Embodiments can provide a computer implemented method for harvesting training data for a training set for use by a system capable of answering questions, the system comprising a processor and a memory comprising instructions executed by the processor, the method comprising receiving, from a user, an input question; processing the input question and returning, to the user, a result set comprising one or more ranked hypotheses and one or more ranked evidence passages corresponding to the one or more ranked hypotheses; receiving, from the user, an indication that one of the one or more ranked hypotheses is to be designated a watched hypothesis; adding the input question and the watched hypothesis to a to-be-vetted question/answer (QA) pair set comprising one or more to-be-vetted QA pairs; vetting each of the one or more to-be-vetted QA pairs in the to-be-vetted QA pair set through a first-pass automatic vetting procedure; if a vetted QA pair passes the first-pass automatic vetting procedure, adding the vetted QA

Abstract: A machine learning hardware accelerator architecture and associated techniques are disclosed. The architecture features multiple memory banks of very wide SRAM that may be concurrently accessed by a large number of parallel operational units. Each operational unit supports an instruction set specific to machine learning, including optimizations for performing tensor operations and convolutions. Optimized addressing, an optimized shift reader and variations on a multicast network that permutes and copies data and associates with an operational unit that support those operations are also disclosed.

Abstract: A non-transitory computer-readable storage medium storing a program that causes a computer to execute a process, the process includes learning a learning model using a first training data group obtained by excluding a test data group from a plurality of training data items; calculating prediction accuracy of the learning model using the test data group; and when the prediction accuracy satisfies the predetermined requirement, learning an error prediction model for determining whether an error of a value predicted by the learning model satisfies a predetermined requirement, by using a second training data group obtained by excluding the test data group and the first training data group from the plurality of training data items.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing inputs using recurrent neural networks. One of the systems includes a main recurrent neural network comprising one or more recurrent neural network layers and a respective hyper recurrent neural network corresponding to each of the one or more recurrent neural network layers, wherein each hyper recurrent neural network is configured to, at each of a plurality of time steps: process the layer input at the time step to the corresponding recurrent neural network layer, the current layer hidden state of the corresponding recurrent neural network layer, and a current hypernetwork hidden state of the hyper recurrent neural network to generate an updated hypernetwork hidden state.

Abstract: A device, method, and non-transitory computer readable storage medium for labelling motion data are provided. The device receives several motion signals, wherein each motion signal includes a motion time message and a motion data group. A motion script includes a plurality of preset motion messages, wherein each preset motion message includes a preset time message and a preset motion. The device performs the following steps for each preset time message: determining a first subset of the motion signals by comparing the motion time messages with the preset time message, calculating a similarity between the motion data group of each motion signal in the first subset and a reference model, determining a second subset of the first subset based on the first similarities, and labelling the motion data group of each motion signal included in the second subset as corresponding to the preset motion corresponding to the preset time message.

Abstract: A system and method for tuning hyperparameters and training a model includes implementing a hyperparameter tuning service that tunes hyperparameters of a model that includes receiving, via an API, a tuning request that includes: (i) a first part comprising tuning parameters for generating tuned hyperparameter values for hyperparameters of the model; and (ii) a second part comprising model training control parameters for monitoring and controlling a training of the model, wherein the model training control parameters include criteria for generating instructions for curtailing a training run of the model; monitoring the training run for training the model based on the second part of the tuning request, wherein the monitoring of the training run includes periodically collecting training run data; and computing an advanced training curtailment instruction based on the training run data that automatically curtails the training run prior to a predefined maximum training schedule of the training run.

Abstract: In an example embodiment, a solution that creates a questionnaire mapping record for questions in a computerized document is utilized to map questions in the computerized document to normalized questions. Where necessary, normalized questions can be automatically created and included in the questionnaire mapping record. Handing strategy rules may also be automatically created for the normalized question, with the handling strategy rules defining how data may be automatically retrieved and used to prepopulate answers to the questions in the computerized document.