pmoe: Penalized Mixtures of Experts
In schiffner/pmoe: Penalized Mixtures of Experts

Description Usage Arguments Value References

Train a penalized mixture of experts model by IPOPT.

pmoe(X, ...)

## S3 method for class 'formula'
pmoe(formula, data, subset, na.action, ...)

## Default S3 method:
pmoe(X, y, colsGating = 1:ncol(X),
  colsExperts = 1:ncol(X), interceptGating = TRUE,
  interceptExperts = TRUE, offsetGating = NULL, offsetExperts = NULL,
  J = 2, lambda, alpha = 1, penalty = c("ungrouped", "grouped"),
  type.multinomial = c("ungrouped", "grouped"), model = c("binomial",
  "multinomial"), standardize = FALSE, genetic = FALSE,
  ipopt.max.iter = 500, ipopt.tol = 1e-06, ...)

`X`	(Required if no `formula` is given as principal argument.) A `matrix` or `data.frame` or `Matrix` containing the explanatory variables (must not contain an intercept column).
`formula`	A `formula` with two right-hand sides of the form `y ~ expert model \| gating model`. The normal formula syntax applies (see `Formula`). In order to use all available columns one or both right-hand sides can be dots (`y ~ . \| .`, `y ~ . \| gating model`). Both right-hand sides normally contain an intercept (which is not penalized) that can be removed by adding `0` or `- 1`. Offsets are allowed, too. The left-hand side should be a vector indicating the class membership, preferably a `factor`.
`data`	A `data.frame` from which variables specified in `formula` are to be taken.
`subset`	A subset...
`na.action`	...
`y`	(Required if no `formula` is given as principal argument.) A `factor` specifying the class membership for each observation.
`colsGating`	Names or indices of columns in `X` to be used for the gating model. Default is all columns.
`colsExperts`	Names or indices of columns in `X` to be used for the expert models. Default is all columns.
`interceptGating`	Logical. Does the gating model include an intercept? If `TRUE`, an intercept column is added to `X`. Defaults to `TRUE`.
`interceptExperts`	Logical. Does the expert model include an intercept? If `TRUE`, an intercept column is added to `X`. Defaults to `TRUE`.
`offsetGating`	Offset term for the gating model.
`offsetExperts`	Offset term for the expert model.
`J`	The number of experts / mixture components. Defaults to 2.
`lambda`	Penalty parameter. Can be a scalar or a vector of length `1+J` with different components for the gating and the `J` expert models. All components must be >= 0.
`alpha`	Mixing parameter for the elastic net penalty. Can be a scalar or a vector of length `1+J` with different components for the gating and the `J` expert models. All components must be in [0,1]. Defaults to 1.
`penalty`	...
`type.multinomial`	`"grouped"`, `"ungrouped"`.
`model`	...
`standardize`	Logical. Should the columns of `X` be standardized prior to fitting the model? Defaults to `FALSE`. If `TRUE` the coefficients are returned on the original scale.
`genetic`	Logical.
`ipopt.max.iter`	The maximum number of IPOPT iterations.
`ipopt.tol`	Tolerance for IPOPT convergence.
`...`	Further arguments.

An object of class pmoe.

Waechter, A. and Biegler, L. T. (2006), On the Implementation of an Interior-Point Filter Line-Search Algorithm for Large-Scale Nonlinear Programming, Mathematical Programming, 106, 25-57.

schiffner/pmoe documentation built on May 29, 2019, 3:39 p.m.