objectiveFast

Computes the objective function for NPMR more quickly than <code>objective</code> by
leveraging pre-computed fitted values, which is the bottleneck in
computing the objective. Intended for internal use only.

Fit multinomial logistic regression with a penalty on the nuclear
norm of the estimated regression coefficient matrix, using proximal
gradient descent.

Scott Powers

npmr

Nuclear Penalized Multinomial Regression

objectiveFast function

<dl> <dt>B</dt>
<dd>fitted regression coefficient matrix</dd> <dt>P</dt>
<dd>matrix of fitted multinomial class probabilities</dd> <dt>W</dt>
<dd>vector containing indices of <code>P</code> which correspond to observed data</dd> <dt>lambda</dt>
<dd>regularization parameter (maybe be a vector of values)</dd></dl>

Arguments

Scott Powers, Trevor Hastie, Rob Tibshirani

Author

Shortcut computation for NPMR objective function — objectiveFast

<dl>

 <dt>B</dt>
<dd>fitted regression coefficient matrix</dd>

 <dt>P</dt>
<dd>matrix of fitted multinomial class probabilities</dd>

 <dt>W</dt>
<dd>vector containing indices of <code>P</code> which correspond to observed data</dd>

 <dt>lambda</dt>
<dd>regularization parameter (maybe be a vector of values)</dd>

</dl>

objectiveFast: Shortcut computation for NPMR objective function

Description

Usage

Value

Arguments

Author

References

See Also