Abstract: Methods, systems, and computer readable storage media are disclosed for generating joint-probabilistic ensemble forecasts for future events based on a plurality of different prediction models for the future events. For example, in one or more embodiments the disclosed system determines error values for various predictions from a plurality of different prediction models (i.e., “forecasters”) for previous events. Moreover, in one or more embodiments the system generates an error probability density function by mapping the error values to an error space and applying a kernel density estimation. Furthermore, the system can apply the error probability density function(s) to a plurality of predictions from the forecasters for a future event to generate a likelihood function and a new prediction for the future event.

Abstract: A proxy-selection system includes an expert rule module that stores a respective set of expert rules applicable to each target device, each expert rule having a respective selection-weight factor. The intelligence module includes logic to acquire, from the separator, the categorized device data for each target device; logic to search, from the expert rule module, a respective set of expert rules which are all applicable to each target device; logic to apply the respective set of expert rules searched for each target device; logic to calculate the respective selection-weight factors of all of the respective set of expert rules applied; and logic to select one or more proxy computers based on a result of the calculation of the selection-weight factors.

Abstract: Techniques are disclosed herein for selecting a predictive model to perform on a set of hierarchical data. A selection of first time series data representing activity observed in a current period in a first hierarchy is received. Second time series data representing activity observed in a prior period in the first hierarchy is retrieved. Predictive models are performed using the second time series data as input, where each predictive model generates statistical outcomes for the current period. A score is generated in each of the plurality of predictive models based on a comparison of the statistical outcomes with the first time series data. Each of the predictive models is ranked based on the generated scores.

Abstract: A preprocessing module of a neural network has a first input and second input. The module generates multiple, different first latent vector representations of its first input, and multiple, different second latent vector representations of its second input. The module then models pairwise interactions between every unique pairwise combination of the first and second latent vector representations. The module then produces an intermediate output by combining the results of the modeled pairwise interactions.

Abstract: In one embodiment, a method includes a preprocessing stage of a neural network model, where the preprocessing stage includes first and second preprocessing modules. Each of the two modules has first input that may receive a dense input and a second input that may receive a sparse input. Each module generates latent vector representations of their respective first and second inputs, and combine the latent vectors with the original first input to define an intermediate output. The intermediate output of the first module is fed into the first input of the second module.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a predictive analytics system that provides a mechanism to change the design or implementations of a product manufactured in a supply chain are disclosed. In one aspect, a method includes the actions of receiving training data that includes private information for a node in a supply chain network and information regarding previous decisions related to product change requests for a product manufactured through the supply chain network; training, using the training data, a predictive model configured to render decisions for requests to change a part used in manufacturing the product; receiving a request to change a given part; applying the predictive model to the request to change the given part; determining a decision approving or denying the request; and transmitting the decision to the requesting node.

Abstract: Tuned hyperparameter values are determined for training a machine learning model. When a selected hyperparameter configuration does not satisfy a linear constraint, if a projection of the selected hyperparameter configuration is included in a first cache that stores previously computed projections is determined. When the projection is included in the first cache, the projection is extracted from the first cache using the selected hyperparameter configuration, and the selected hyperparameter configuration is replaced with the extracted projection in the plurality of hyperparameter configurations. When the projection is not included in the first cache, a projection computation for the selected hyperparameter configuration is assigned to a session. A computed projection is received from the session for the selected hyperparameter configuration.

Abstract: A random binning featurization process method, system, and computer program product for a distributed random binning featurization process on one or more multicore systems with a hybrid two-level parallelism, the method including in a training phase, receiving a first data matrix dividing the random binning featurization process into two orthogonal levels, in a high-level generating a randomized number of high-dimension grids and evenly partitioning the grids into nodes in a parallel system, and in a low-level, evenly partitioning dimensions in each grid to construct look-up tables of index vectors and compute a local feature matrix for each node.

Abstract: A method and system for improving a stochastic control problem policy, the method including a sampling device obtaining data representing sample Boltzmann machine configurations, obtaining a stochastic control problem's initialization data and initial policy; assigning representative data of initial coupler weights and node biases and the Boltzmann machine's transverse field strength to the sampling device; until a stopping criterion is met, generating a present-epoch state-action pair, amending, sampling for the present-epoch state-action pair, approximating a present-epoch state-action Q-function value, obtaining a future-epoch state-action pair through a stochastic state process including a stochastic optimization test on all state-action pairs to provide the action at the future-epoch and update the future-epoch state's policy; amending the representative data, sampling for the future-epoch state-action pair, obtaining a future-epoch state-action Q-function value, updating each weight and bias and providi

Abstract: The present disclosure relates to an asymmetric font pairing system that efficiently pairs digital fonts. For example, in one or more embodiments, the asymmetric font pairing system automatically identifies and provides users with visually aesthetic font pairs for use in different sections of an electronic document. In particular, the asymmetric font pairing system learns visually aesthetic font pairs using joint symmetric and asymmetric compatibility metric learning. In addition, the asymmetric font pairing system provides compact compatibility spaces (e.g., a symmetric compatibility space and an asymmetric compatibility space) to computing devices (e.g., client devices and server devices), which enable the computing devices to quickly and efficiently provide font pairs to users.

Abstract: A system and method of configuring a graphical control structure for controlling a machine learning-based automated dialogue system includes configuring a root dialogue classification node that performs a dialogue intent classification task for utterance data input; configuring a plurality of distinct dialogue state classification nodes that are arranged downstream of the root dialogue classification node; configuring a graphical edge connection between the root dialogue classification node and the plurality of distinct state dialogue classification nodes that graphically connects each of the plurality of distinct state dialogue classification nodes to the root dialogue classification node, wherein (i) the root dialogue classification node, (ii) the plurality of distinct classification nodes, (iii) and the transition edge connections define a graphical dialogue system control structure that governs an active dialogue between a user and the machine learning-based automated dialogue system.

Abstract: Disclosed is a method and a device of trajectory outlier detection. The method may include: points on a trajectory to be detected are obtained by sampling the trajectory; characteristic points are extracted from the points according to spatial state and temporal state of each of the points; trajectory segments are obtained by segmenting the trajectory according to the characteristic points; each of the trajectory segments is compared to normal trajectory segments and abnormal trajectory segments; and one or more trajectory outliers are identified from the trajectory segments based on comparison results. Wherein, the normal trajectory segments and the abnormal trajectory segments are obtained by clustering trajectory segments in a training set; and the trajectory segments in the training set are obtained by segmenting historical trajectories based on characteristic points extracted from points on the historical trajectories according to spatial state and temporal state of the points.