compose_regularization: Compose an existing loss function with regularization

In holub008/gdlm: Build linear models through gradient descent

Description

Given an arbitrary loss function, compose it with elastic net regularization

Usage

compose_regularization(loss, elastic_net_parameter, lambda)

Arguments

loss	the loss function to be built on
elastic_net_parameter	the weighting of L1 regularization in the elastic net. L2 will be weighted as 1 - elastic_net_parameter
lambda	parameter controlling how much regularization contributes to the overall loss

Author(s)

kholub

Examples

lasso_ls <- compose_regularization(LS_LOSS, 1, 1e-5)
ridge_ls <- compose_regularization(LS_LOSS, 0, 1e-5)

holub008/gdlm documentation built on May 13, 2019, 6:14 p.m.