Oracle International Corporation

Patrick Aboyoun

Vaishnavi A Sashikanth

Dmitry Golovashkin

Denis B Mukhin

Lei Zhang

Computer systems, machine-implemented methods, and stored instructions are provided for minimizing an approximate global error in an artificial neural network that is configured to predict model outputs based at least in part on one or more model inputs. A model manager stores the artificial neural network model. The model manager may then minimize an approximate global error in the artificial neural network model at least in part by causing evaluation of a mixed integer linear program that determines weights between artificial neurons in the artificial neural network model. The mixed integer linear program accounts for piecewise linear activation functions for artificial neurons in the artificial neural network model. The mixed integer linear program comprises a functional expression of a difference between actual data and modeled data, and a set of one or more constraints that reference variables in the functional expression.

Minimizing global error in an artificial neural network

A system, and computer program product for large-scale data transformations. Embodiments include a smoothing engine within an R environment to configure at least one master task and at least two worker tasks. A chunk calculator receives a series of data values and divides the series of data values into portions of data values which are in turn assigned as workloads to at least two worker tasks. The worker tasks serve to calculate a first state value of a first one of the portions of data values, and calculate a second state value of a second one of the portions of data values. The workloads are selected such that calculating a second state value does not depend on the first state value. The results of the workload calculations are used to calculate a smoothing factor used to predict a trend.

Exploiting parallelism in exponential smoothing of large-scale discrete datasets

A method, system, and computer program product for interfacing an R language client with a separate database engine environment. The method commences by interpreting an R language code fragment to identify and select R language constructs and transforming the R language constructs into queries or other database language constructs to execute within the database engine environment. The method further implements techniques for transmitting marshalled results (resulting from the execution of the database language constructs) back to the R client environment. In some situations, the marshalled results include an XML schema or DTD or another metadata description of the structure of the results. User conveniences in the form of transparent database query construction, and transparent exploitation of parallelism is achieved by determining the computation resource requirements of executable language constructs, and mapping the execution to one (or more) of, a local R client, a database engine, or a remote execution engine.

Overloading R language constructs with database engine constructs

According to one aspect of the invention, target data comprising observations is received. A neural network comprising input neurons, output neurons, hidden neurons, skip-layer connections, and non-skip-layer connections is used to analyze the target data based on an overall objective function that comprises a linear regression part, the neural network's unregularized objective function, and a regularization term. An overall optimized first vector value of a first vector and an overall optimized second vector value of a second vector are determined based on the target data and the overall objective function. The first vector comprises skip-layer weights for the skip-layer connections and output neuron biases, whereas the second vector comprises non-skip-layer weights for the non-skip-layer connections.

Growing community of inventors

Seattle, WA, United States of America

Patrick Aboyoun