MKLR

Hi,guys! Welcome to MKLP package homepage. I used the knowledge about how to create a R package in BIS620 to solve a problem in BIS555.

Multi-class kernel logistic regression. This function fit a Multi-class Kernel Logistic Regression model to the data. The return list contains the estimated kernel parameters and logistic parameters.There are two types of kernel, they are RBF and polynomial.

I refer to KLR and the Multinomial Kernel Logistic Regression via Bound Optimization Approach for the calculation and use gradient descent (not dual or fix the parameter) to get all parameters.

And I will keep updating the package and try to provide methods like CV, LOOCV and more for better model selection.

Installation

You can install the released version of bubblematrix from GitHub with:

# install.packages("devtools")
library(devtools)
devtools::install_github("hyj12345/MKLR")

Example

I use the data for my course homework as a simple example.

View the built-in dataset

library(readr)
library(magrittr)
train_data <- read_csv("~/Desktop/21Fall/BIS555/vowel/training")%>%.[2:12]
test_data <- read_csv("~/Desktop/21Fall/BIS555/vowel/test")%>%.[2:12]

• Train

library(MKLR)
##Train the model
model_mklr<-MKLR(train_data$y,train_data[,-1],max_iter=1000,threshold=1.0e-5,lr=0.5,kernel = 'RBF')

• Predict

return classes or probabilities

pre_mklr<-MKLR::predict.MKLR(model_mklr,test_data[,-1],response = 'class')