Sas Institute Inc.

Ruiwen Zhang

Robert Winston Blanchard

Jorge Manuel Gomes Da Silva

Jun Liu

Prathaban Mookiah

Jared Langford Dean

Weichen Wang

Yongqiao Xiao

Taiyeong Lee

Hamoon Azizsoltani

Ye Liu

Joshua David Griffin

Yan Xu

Varunraj Valsaraj

Samuel Paul Leeman-Munk

Mohammadreza Nazari

Zheng Zhao

Jonathan Lee Walker

Amirhassan Fallah Dizche

William Robert Nadolski

Hardi Desai

Davood Hajinezhad

Afshin Oroojlooyjadid

Rui Liu

Bingfeng (Ben) Ding

Lochan Basnet

Vrushali Sawant

A system and method include generating synthetic data by generating a first set of hyperparameters for a first trained machine learning model and a second set of hyperparameters for a second trained machine learning model, generating a plurality of synthetic data vectors using the first and second trained machine learning models, computing an error function for the first and second set of hyperparameters using a third machine learning model, computing an objective function value, responsive to determining that the objective function value is not an optimal value, updating the first set of hyperparameters and the second set of hyperparameters or responsive to determining that the objective function value is an optimal value outputting the plurality of synthetic data vectors as a set of synthetic data.

Techniques for generating synthetic data

A system and method include forecasting a series of future data points in a long sequence time series data using a decoder-only transformer model by dividing the long sequence time series data into a plurality of sequences, converting each sequence of the plurality of sequences into a first vector to obtain a plurality of first vectors, creating a plurality of second vectors from the time stamps associated with the plurality of data points, combining the first vector with the second vector of each sequence of the plurality of sequences to obtain a plurality of third vectors, computing a context matrix from the plurality of third vectors, performing a convolution operation on the context matrix to forecast the series of future data points, and outputting the series of future data points from the prediction layer.

Decoder-only transformer model for time series data

A computing device trains a machine state predictive model. A generative adversarial network with an autoencoder is trained using a first plurality of observation vectors. Each observation vector of the first plurality of observation vectors includes state variable values for state variables and an action variable value for an action variable. The state variables define a machine state, wherein the action variable defines a next action taken in response to the machine state. The first plurality of observation vectors successively defines sequential machine states to manufacture a product. A second plurality of observation vectors is generated using the trained generative adversarial network with the autoencoder. A machine state machine learning model is trained to predict a subsequent machine state using the first plurality of observation vectors and the generated second plurality of observation vectors. A description of the machine state machine learning model is output.

Automated control of a manufacturing process

(A) Conditional vectors are defined. (B) Latent observation vectors are generated using a predefined noise distribution function. (C) A forward propagation of a generator model is executed with the conditional vectors and the latent observation vectors as input to generate an output vector. (D) A forward propagation of a decoder model of a trained autoencoder model is executed with the generated output vector as input to generate a plurality of decoded vectors. (E) Transformed observation vectors are selected from transformed data based on the defined plurality of conditional vectors. (F) A forward propagation of a discriminator model is executed with the transformed observation vectors, the conditional vectors, and the decoded vectors as input to predict whether each transformed observation vector and each decoded vector is real or fake. (G) The discriminator and generator models are updated and (A) through (G) are repeated until training is complete.

Tabular data generation for machine learning model training system

Processing speeds for generating a model can be enhanced. For example, the model can be generated by using regression coefficient values as weights for independent variables in the model. The regression coefficient values can be determined using a coordinate descent method to find a minimum value of a least absolute shrinkage and selection operator cost function. Each iteration of the coordinate descent method can include determining a starting coordinate based on (i) a previous starting coordinate or a previous regression coefficient value from an immediately prior iteration of the coordinate descent method; (ii) a current regression coefficient value associated with a current iteration of the coordinate descent method; and (iii) a refinement factor configured to minimize a result of a univariate algorithm. Each iteration can also include performing a coordinate descent using the starting coordinate to determine a next regression coefficient value for a next iteration of the coordinate descent method.

Enhancing processing speeds for generating a model on an electronic device

A method of converting data to tree data is provided. A first node memory structure that includes a first value indicator, a first counter value, and a first observation indicator is initialized for a first variable. The first value indicator is initialized with a first value of the first variable selected from first observation data, and the first observation indicator is initialized with a first indicator that indicates the first observation data. The first value of the first variable is compared to a second value of the first variable. The first counter value included in the first node memory structure is incremented when the first value of the first variable matches the second value of the first variable. Corresponding values of second observation data are compared to the identified values from first observation data when the first value of the first variable matches the second value of the first variable. A next observation is read from the data when the identified values match the corresponding values. The tree data is output after a last observation of the data is processed.

Data structure supporting contingency table generation

A method of creating a contingency table is provided. Whether or not a variable level list exists for a second variable in tree data is determined. When the variable level list exists for the second variable in the tree data, a first node memory structure is determined for the second variable from the variable level list, a first value of a first variable is determined using a first observation indicator and the tree data, and a first counter value is added to the contingency table in association with the first value of the first variable and a first value of the second variable. The first node memory structure includes the first value indicator, the first counter value, and the first observation indicator. The first value indicator indicates a first value of the second variable.

Contingency table generation

A power method can be enhanced. For example, an electronic communication indicating a job to be performed can be received. A best rank-1 approximation of a matrix associated with the job can be determined using the power method. Each iteration of the power method can include determining a point that lies on a line passing through (i) a first value for a first singular vector from an immediately prior iteration of the power method; and (ii) a second value for the first singular vector from another prior iteration of the power method. Each iteration of the power method can also include determining, by performing the power method using the point, a current value for the first singular vector and a current value for a second singular vector for a current iteration of the power method. The job can then be performed using the best rank-1 approximation of the matrix.

Enhanced power method on an electronic device

Growing community of inventors

Cary, NC, United States of America

Ruiwen Zhang