loss2_ccsvm: Composite Loss Value for epsilon-insensitive Type
In mpath: Regularized Linear Models

loss2_irsvm

R Documentation

Composite Loss Value for epsilon-insensitive Type

Compute composite loss value for epsilon-insensitive type function

loss2_irsvm(y, f, weights, cfun, dfun, s, eps, delta=0.0001)

`y`	response variable values
`f`	fitted values of `y`
`weights`	observation weights, same length as `y`
`cfun`	integer from 1-8, concave function as in `irsvm_fit`
`dfun`	integer value, only `dfun=2` is implemented for now. Convex function as in `irsvm_fit`
`s`	tuning parameter of `cfun`. `s > 0` and can be equal to 0 for `cfun="tcave"`.
`delta`	a small positive number provided by user only if `cfun="gcave"` and `0 < s <1`
`eps`	non-negative parameter for epsilon-insensitive loss

For large s values, the loss can be 0 with cfun=2,3,4, or "acave", "bcave", "ccave".

Weighted loss values

Zhu Wang <zwang145@uthsc.edu>

Zhu Wang (2020) Unified Robust Estimation, arXiv e-prints, https://arxiv.org/abs/2010.02848