mxComputeGradientDescent: Optimize parameters using a gradient descent optimizer

In OpenMx: Extended Structural Equation Modelling

Optimize parameters using a gradient descent optimizer

Description

This optimizer does not require analytic derivatives of the fit function. The fully open-source CRAN version of OpenMx offers 2 choices, CSOLNP and SLSQP (from the NLOPT collection). The OpenMx Team's version of OpenMx offers the choice of three optimizers: CSOLNP, SLSQP, and NPSOL.

Usage

mxComputeGradientDescent(
  freeSet = NA_character_,
  ...,
  engine = NULL,
  fitfunction = "fitfunction",
  verbose = 0L,
  tolerance = NA_real_,
  useGradient = deprecated(),
  warmStart = NULL,
  nudgeZeroStarts = mxOption(NULL, "Nudge zero starts"),
  maxMajorIter = NULL,
  gradientAlgo = deprecated(),
  gradientIterations = deprecated(),
  gradientStepSize = deprecated()
)

Arguments

freeSet	names of matrices containing free parameters.
...	Not used. Forces remaining arguments to be specified by name.
engine	specific 'CSOLNP', 'SLSQP', or 'NPSOL'
fitfunction	name of the fitfunction (defaults to 'fitfunction')
verbose	integer. Level of run-time diagnostic output. Set to zero to disable
tolerance	how close to the optimum is close enough (also known as the optimality tolerance)
useGradient	\lifecycle soft-deprecated
warmStart	a Cholesky factored Hessian to use as the NPSOL Hessian starting value (preconditioner)
nudgeZeroStarts	whether to nudge any zero starting values prior to optimization (default TRUE)
maxMajorIter	maximum number of major iterations
gradientAlgo	\lifecycle soft-deprecated
gradientIterations	\lifecycle soft-deprecated
gradientStepSize	\lifecycle soft-deprecated

Details

All three optimizers can use analytic gradients, and only NPSOL uses warmStart. To customize more options, see mxOption.

References

Luenberger, D. G. & Ye, Y. (2008). Linear and nonlinear programming. Springer.

Examples

data(demoOneFactor)
factorModel <- mxModel(name ="One Factor",
  mxMatrix(type="Full", nrow=5, ncol=1, free=FALSE, values=0.2, name="A"),
    mxMatrix(type="Symm", nrow=1, ncol=1, free=FALSE, values=1, name="L"),
    mxMatrix(type="Diag", nrow=5, ncol=5, free=TRUE, values=1, name="U"),
    mxAlgebra(expression=A %*% L %*% t(A) + U, name="R"),
  mxExpectationNormal(covariance="R", dimnames=names(demoOneFactor)),
  mxFitFunctionML(),
    mxData(observed=cov(demoOneFactor), type="cov", numObs=500),
     mxComputeSequence(steps=list(
     mxComputeGradientDescent(),
     mxComputeNumericDeriv(),
     mxComputeStandardError(),
     mxComputeHessianQuality()
    )))
factorModelFit <- mxRun(factorModel)
factorModelFit$output$conditionNumber # 29.5

OpenMx documentation built on Oct. 19, 2024, 9:06 a.m.