README.md

In JamilWaqas/CR: Competitive Regression

Proto-type Algorithms

• AAR
• ORR
• COIRR
• OSLOG

Benchmarks-glmnet

• lm regression
• L1 regression
• L2 regression

Grid search tuning

• tuneAAR
• tuneORR
• tuneCOIRR
• tuneOSLOG

Time-microbenchmark

• time

Note:- Algorithms are not optimised for performance. If inverse is taken using solve(), then ORR and AAR are quite slower than COIRR and OSLOG. When chol2inv(chol()) is used than they are faster than OSLOG and COIRR. However, COIRR and OSLOG has ability to give sparse solution. Also, there might be room for further optimising multiplication that involve diagonal matrices. Other techniques like Sherman-Morrison can also be used. The current implementations are just proto-types, which possibly also effect the computationally efficiency. These aspect are still work in progress, so please feel to improve the current implementations.

Instructions

Open R and install the following packages microbenchmark, caret, glmnet and devtools by typing install.packages("pkg.name") followed by library(pkg.name). Then type inatall_gthub("Jamil/CR")

Example

Type source("https://raw.githubusercontent.com/JamilWaqas/CR/master/demo.R") in R.

JamilWaqas/CR documentation built on Sept. 7, 2020, 7:34 p.m.