NEWS.md

In lukesonnet/KRLS: Kernel-based Regularized Least squares

KRLS 1.1.0 (Current Github version)

• Temporarily rename package KRLS2
• Use C++ for speed improvements throughout
• Add ability to use logistic loss instead of least squares loss for binary outcomes (paper forthcoming)
• Enable truncation to improve speed with large datsets
• Change sigma to b for consistency
• Change default value of b from p to 2*p where p is the number of covariates
• Change first differences algorithm for speed, including correcting variance of average first differences, which was too large by a factor of 2

KRLS 1.0.0 (Current CRAN version)

• Main functionality as described in Hainmueller and Hazlett (2014) [ungated]

lukesonnet/KRLS documentation built on May 21, 2019, 8:58 a.m.