Sas Institute Inc.

Jan Chvosta

Mark Roland Little

Xilong Chen

Christian Macaro

Mahesh V Joshi

Richard Potter

Gunce Eryuruk Walton

Tao Huang

Douglas Allan Cairns

David Bruce Elsheimer

Jared Langford Dean

Xunlei Wu

Ming-Chun Chang

Yang Zhao

Michael Thomas Lamm

Ji Shen

Guohui Wu

Wan Xu

A computing device learns a best topological order vector of a plurality of variables. A target variable and zero or more input variables are defined. (A) A machine learning model is trained with observation vectors using the target variable and the zero or more input variables. (B) The machine learning model is executed to compute an equation loss value. (C) The equation loss value is stored with the identifier. (D) The identifier is incremented. (E) (A) through (D) are repeated a plurality of times. (F) A topological order vector is defined. (G) A loss value is computed from a subset of the stored equation loss values based on the topological order vector. (F) through (G) are repeated for each unique permutation of the topological order vector. A best topological order vector is determined based on a comparison between the loss value computed for each topological order vector in (G).

Topological order determination using machine learning

A computer system can automatically analyze a video of a physical activity and provide corresponding feedback. For example, the system can receive a video file including image frames showing an entity performing a physical activity that involves a sequence of movement phases. The system can generate coordinate sets by performing image analysis on the image frames. The system can provide the coordinate sets as input to a trained model, the trained model being configured to assign scores and movement phases to the image frames based on the coordinate sets. The system can then select a particular movement phase for which to provide feedback, based on the scores and movement phases assigned to the image frames. The system can generate the feedback for the entity about their performance of the particular movement phase, which may improve the entity's future performance of that particular movement phase.

Computer system for automatically analyzing a video of a physical activity using a model and providing corresponding feedback

A computing device learns a directed acyclic graph (DAG). An SSCP matrix is computed from variable values defined for observation vectors. A topological order vector is initialized that defines a topological order for the variables. A loss value is computed using the topological order vector and the SSCP matrix. (A) A neighbor determination method is selected. (B) A next topological order vector is determined relative to the initialized topological order vector using the neighbor determination method. (C) A loss value is computed using the next topological order vector and the SSCP matrix. (D) (B) and (C) are repeated until each topological order vector is determined in (B) based on the neighbor determination method. A best topological vector is determined from each next topological order vector based on having a minimum value for the computed loss value. An adjacency matrix is computed using the best topological vector and the SSCP matrix.

Directed acyclic graph machine learning system

A treatment model that is a first neural network is trained to optimize a treatment loss function based on a treatment variable t using a plurality of observation vectors by regressing t on x,z. The trained treatment model is executed to compute an estimated treatment variable value {circumflex over (t)}for each observation vector. An outcome model that is a second neural network is trained to optimize an outcome loss function by regressing y on xand an estimated treatment variable t. The trained outcome model is executed to compute an estimated first unknown function value {circumflex over (&#x3b1;)}(x) and an estimated second unknown function value {circumflex over (&#x3b2;)}(x) for each observation vector. An influence function value is computed for a parameter of interest using {circumflex over (&#x3b1;)}(x) and {circumflex over (&#x3b2;)}(x). A value is computed for the predefined parameter of interest using the computed influence function value.

Causal inference and policy optimization system based on deep learning models

A computing device selects a trained spatial regression model. A spatial weights matrix defined for observation vectors is selected, where each element of the spatial weights matrix indicates an amount of influence between respective pairs of observation vectors. Each observation vector is spatially referenced. A spatial regression model is selected from spatial regression models, initialized, and trained using the observation vectors and the spatial weights matrix to fit a response variable using regressor variables. Each observation vector includes a response value for the response variable and a regressor value for each regressor variable of the regressor variables. A fit criterion value is computed for the spatial regression model and the spatial regression model selection, initialization, and training are repeated until each spatial regression model is selected. A best spatial regression model is selected and output as the spatial regression model having an extremum value of the fit criterion value.

Automatic spatial regression system

Computing resources consumed in performing computerized sequence-mining can be reduced by implementing some examples of the present disclosure. In one example, a system can determine weights for data entries in a data set and then select a group of data entries from the data set based on the weights. Next, the system can determine a group of k-length sequences present in the selected group of data entries by applying a shuffling algorithm. The system can then determine frequencies corresponding to the group of k-length sequences and select candidate sequences from among the group of k-length sequences based on the frequencies thereof. Next, the system can determine support values corresponding to the candidate sequences and then select output sequences from among the candidate sequences based on the support values thereof. The system may then transmit an output signal indicating the selected output sequences an electronic device.

Reducing consumption of computing resources in performing computerized sequence-mining on large data sets

Techniques for automated Bayesian posterior sampling using Markov Chain Monte Carlo and related schemes are described. In an embodiment, one or more values in a stationarity phase for a system configured for Bayesian sampling may be initialized. Sampling may be performed in the stationarity phase based upon the one or more values to generate a plurality of samples. The plurality of samples may be evaluated based upon one or more stationarity criteria. The stationarity phase may be exited when the plurality of samples meets the one or more stationarity criteria. Other embodiments are described and claimed.

Techniques for automated Bayesian posterior sampling using Markov Chain Monte Carlo and related schemes

Techniques for estimated compound probability distribution are described herein. Embodiments may include receiving a compound model specification comprising a frequency model and a severity model, the compound model specification including a model error comprising a frequency model error and a severity model error, and determining a number of frequency models and severity models to generate based on the received number of models to generate. Embodiments include generating a plurality of frequency models through perturbation of the frequency model according to the frequency model error, and generating a plurality of severity models through perturbation of the severity model according to the severity model error. Further, embodiments include dividing generation of a plurality of compound model samples among a plurality of distributed worker nodes, and receiving the plurality of compound model samples from the distributed worker nodes, and generating aggregate statistics from the plurality of compound model samples.

Techniques for estimating compound probability distribution by simulating large empirical samples with scalable parallel and distributed processing

Various embodiments are directed to techniques for deriving a sample representation from a random sample. A computer-program product includes instructions to cause a first computing device to fit an empirical distribution function to a marginal probability distribution of a variable within a first sample portion of a random sample to derive a partial marginal probability distribution approximation, wherein the random sample is divided into multiple sample portions distributed among multiple computing devices; fit a first portion of a copula function to a multivariate probability distribution of the first sample portion, wherein the copula function is divided into multiple portions; and transmit an indication of a first likelihood contribution of the first sample portion to a coordinating device to cause a second computing device to fit a second portion of the copula function to a multivariate probability distribution of a second sample portion. Other embodiments are described and claimed.

Approximate multivariate posterior probability distributions from simulated samples

Various embodiments are generally directed to techniques for producing statistically correct and efficient combinations of multiple simulated posterior samples from MCMC and related Bayesian sampling schemes are described. One or more chains from a Bayesian posterior distribution of values may be generated. It may be determine whether the one or more chains have reached stationarity through parallel processing on a plurality of processing nodes. Based upon the determination, each of the one or more chains that have reached stationarity through parallel processing on the plurality of processing nodes may be sorted. The one or more sorted chains may be resampled through parallel processing on the plurality of processing nodes. The one or more resampled chains may be combined. Other embodiments are described and claimed.

Techniques for producing statistically correct and efficient combinations of multiple simulated posterior samples

Techniques for estimated compound probability distribution are described herein. Embodiments may include receiving, at a master node of a distributed system, a compound model specification comprising frequency models, severity models, and one or more adjustment functions, wherein at least one model of the frequency models and the severity models depend on one or more regressor and distributing the compound model specification to worker nodes of the distributed system, each of the worker nodes to at least generate a portion of samples for use in predicting compound distribution model estimates. Embodiments may also include predicting the compound distribution model estimates based on the sample portions of aggregate values and adjusted aggregate values.

Techniques for automated Bayesian posterior sampling using Markov Chain Monte Carlo and related schemes are described. In an embodiment, one or more values in an accuracy phase for a system configured for Bayesian sampling may be initialized. Sampling may be performed in the accuracy phase based upon the one or more values to generate a plurality of samples. The plurality of samples may be evaluated based upon one or more accuracy criteria. The accuracy phase may be exited when the plurality of samples meets the one or more accuracy criteria. Other embodiments are described and claimed.

Techniques for automated bayesian posterior sampling using markov chain monte carlo and related schemes

Various embodiments are directed to techniques for selecting a subset of a set of simulated samples. A computer-program product including instructions to cause a computing device to order a plurality of UPDFs by UPDF value, wherein the plurality of UPDFs is associated with a chain of draws of a set of simulated samples, wherein each draw comprises multiple parameters and the UPDF values map to parameter values of the parameters; select a subset of the plurality of UPDFs based on the subset of the plurality of UPDFs having UPDF values within a range corresponding to a range of parameter values to include in a subset of the set of simulated samples; and transmit an indication of a draw comprising parameters having parameter values to include in the subset of the set of simulated samples, wherein the indication identifies the draw by associated UPDF. Other embodiments are described and claimed.

Compact representation of multivariate posterior probability distribution from simulated samples

Techniques for estimated compound probability distribution are described. An apparatus comprising a configuration component, perturbation component, sample generation controller, an aggregation component, a distribution fitting component, and statistics generation component. The configuration component operative to receive a compound model specification and candidate distribution definition. The perturbation component operative to generate a plurality of models from the compound model specification. The sample generation controller operative to initiate the generation of a plurality of compound model samples from each of the plurality of models. The distribution fitting component to generate parameter values for the candidate distribution definition based on the compound model samples. The statistics generation component to generate approximated aggregate statistics.

Growing community of inventors

Raleigh, NC, United States of America

Jan Chvosta