초록 ▼

A multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. This computing architecture, referred to as a concurrent-learning information proc

A multi-kernel neural network computing architecture configured to learn correlations among feature values 34, 38 as the network monitors and imputes measured input values 30 and also predicts future output values 46. This computing architecture, referred to as a concurrent-learning information processor (CIP 10), includes a multi-kernel neural network array 14 with the capability to learn and predict in real time. The CIP 10 also includes a manager 16 and an input-output transducer 12 that may be used for input-output refinement. These components allow the computing capacity of the multi-kernel array 14 to be reassigned in response to measured performance or other factors. The output feature values 46 computed by the multi-kernel array 14 and processed by an output processor 44 of the transducer 12 are supplied to a response unit 18 that may be configured to perform a variety of monitoring, forecasting, and control operations in response to the computed output values. Important characteristics of the CIP 10, such as feature function specifications 35 and 49, connection specifications 42, learning weight schedules 55, and the like may be set by a technician through a graphical user interface 20. Refinement processes also allow the CIP 10 be reconfigured in accordance with user commands for application to different physical applications.

대표청구항 ▼

[ What is claimed is:] [1.]1. A method for responding to computed output values that are based on measured input values received during a current time trial and during one or more historical time trials, comprising the steps of:(a) receiving an iteration of the measured input values for the current

[ What is claimed is:] [1.]1. A method for responding to computed output values that are based on measured input values received during a current time trial and during one or more historical time trials, comprising the steps of:(a) receiving an iteration of the measured input values for the current time trial;(b) assembling a vector of input feature values based on the measured input values;(c) providing the vector of input feature values to a multi-kernel processor, each kernel of the processor operative for:receiving one or more of the input feature values,retrieving connection specifications defining mathematical relationships for computing one or more output feature values based on the received input feature values,retrieving a set of connection weights reprinting regression coefficients among the received input feature values and the computed output feature values,retrieving a set of learning weights defining mathematical relationships for updating the connection weights based on the received input feature values,computing the output feature values based on the received input feature values, the connection weights, and the connection specifications,computing updated connection weights based on the received input feature values, the connection weights, the connection specifications, and the leaning weights,storing the updated connection weights, andproviding access to the computed output feature values;(d) assembling a vector of computed output values based on the output feature values computed by each kernel;(e) responding to the vector of computed output values;(f) determining whether refinement operations are indicated, and if refinement operations are not indicated, repeating steps (a) through (f) for a subsequent time trial; and(e) if refinement operations are indicated, performing one or more refinement operations selected from the group including:recomputing the learning weights based on the measured input values and the computed output values for a plurality of time trials;recomputing the connection specifications based on the measured input values and the computed output values for a plurality of time trials;deleting ineffective input or output feature values;combining redundant input or output feature values;specifying new input or output feature values;recomputing the input feature specifications based on the measured input values and the computed output values for a plurality of time trials;recomputing the output feature specifications based on the measured input values and the computed output values for a plurality of time trials; andreassigning functionality among the kernels.