×

Performance of artificial neural network models in the presence of instrumental noise and measurement errors

  • US 7,313,550 B2
  • Filed: 03/27/2002
  • Issued: 12/25/2007
  • Est. Priority Date: 03/27/2002
  • Status: Active Grant
First Claim
Patent Images

1. A computer-implemented method for improving the prediction accuracy and generalization performance of nonlinear artificial neural network models when an example set of input-output data, available for constructing the network model, comprises at least one of instrumental noise and measurement errors, said method comprising the steps of:

  • (a) generating noise-superimposed enlarged input-output sample data set using computer simulations;

    (b) generating for each input-output pattern in the example set, M number of noise-superimposed sample input-output patterns using computer simulations;

    (c) generating noise-superimposed sample input-output patterns using noise tolerance values, which are specific to each input-output variable;

    (d) generating Gaussian distributed random numbers using computer simulations to create noise-superimposed sample input-output patterns;

    (e) determining an exact amount of Gaussian noise to be added to each input-output variable in the example set by using a stochastic search and optimization technique; and

    (f) constructing the nonlinear artificial neural network model using a training set, the training set comprising the computer generated noise-superimposed sample input-output patterns,wherein the nonlinear artificial neural network models comprise a model of an industrial polymerization process or a jacketed non-isothermal continuous stirred tank reactor,wherein input-output variables represent at least two of a temperature, a pressure level, a liquid, and a gas, andwherein at least one input variable effects at least one output variable differently from another input variable, and wherein the exact amount of Gaussian noise for each of input variables is determined using the stochastic search and optimization technique based on effect of a respective input variable on said at least one output variable.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×