train_TRMF: Train a TRMF model

In TRMF: Temporally Regularized Matrix Factorization

Train a TRMF model

Description

This function is the "engine" of the TRMF package. It takes a previously created TRMF object, sets up the internal objects and fits it to the data using an alternating least squares (ALS) algorithm.

Usage

## S3 method for class 'TRMF'
train(x,numit=10,Xm=NULL,Fm=NULL,Z=NULL,...)

Arguments

x	A TRMF object to be fit.
numit	Number of alternating least squares iterations
Xm	Optional initial matrix. See details
Fm	Optional initial matrix. See details
Z	Optional initial vector. See details
...	ignored

Details

If a coefficient model is not present in object, it adds a L2 regularization model. If no time series models have been added to object, it adds a simple model using TRMF_simple. If Xm, Fm are both NULL, then initial estimates are provided. If both are provided, the ALS algorithm is started with Xm (with a warning) and Fm is overwritten with the next ALS step. If external regressors are in the model and Fm is provided, the number of rows of columns must equal number of regressors plus columns of Xm. The format of Fm in this case is: [column_xreg_parameters,global_xreg_parameters,Fm_parameters]^T. If Z is NULL, then a vector of 1's is used as initial estimate.

Value

train returns a fitted object of class "TRMF" that contains the data, all added models, matrix factorization and fitted model. The matrix factors Xm, Fm are stored in object$Factors$Xm and object$Factors$Fm, object$Factors$Z respectively. Use fitted to get fitted model, use resid to get residuals, use coef to get coefficients (Fm matrix) and components to get Xm or Fm.

Author(s)

Chad Hammerquist

References

Yu, Hsiang-Fu, Nikhil Rao, and Inderjit S. Dhillon. "High-dimensional time series prediction with missing values." arXiv preprint arXiv:1509.08333 (2015).

See Also

create_TRMF, TRMF_columns, TRMF_trend

Examples

# create test data
xm = poly(x = (-10:10)/10,degree=4)
fm = matrix(rnorm(40),4,10)
Am = xm%*%fm+rnorm(210,0,.2)

# create model
obj = create_TRMF(Am)
out = train(obj)
plot(out)

TRMF documentation built on June 22, 2024, 10:41 a.m.