WellSVMSSLR.Rdmodel from RSSL package WellSVM is a minimax relaxation of the mixed integer programming problem of finding the optimal labels for the unlabeled data in the SVM objective function. This implementation is a translation of the Matlab implementation of Li (2013) into R.
WellSVMSSLR( C1 = 1, C2 = 0.1, gamma = 1, x_center = TRUE, scale = FALSE, use_Xu_for_scaling = FALSE, max_iter = 20 )
| C1 | double; A regularization parameter for labeled data, default 1; |
|---|---|
| C2 | double; A regularization parameter for unlabeled data, default 0.1; |
| gamma | double; Gaussian kernel parameter, i.e., k(x,y) = exp(-gamma^2||x-y||^2/avg) where avg is the average distance among instances; when gamma = 0, linear kernel is used. default gamma = 1; |
| x_center | logical; Should the features be centered? |
| scale | logical; Should the features be normalized? (default: FALSE) |
| use_Xu_for_scaling | logical; whether the unlabeled objects should be used to determine the mean and scaling for the normalization |
| max_iter | integer; Maximum number of iterations |
Y.-F. Li, I. W. Tsang, J. T. Kwok, and Z.-H. Zhou. Scalable and Convex Weakly Labeled SVMs. Journal of Machine Learning Research, 2013.
R.-E. Fan, P.-H. Chen, and C.-J. Lin. Working set selection using second order information for training SVM. Journal of Machine Learning Research 6, 1889-1918, 2005.
library(tidyverse) library(tidymodels) library(caret) library(SSLR) data(breast) set.seed(1) train.index <- createDataPartition(breast$Class, p = .7, list = FALSE) train <- breast[ train.index,] test <- breast[-train.index,] cls <- which(colnames(breast) == "Class") #% LABELED labeled.index <- createDataPartition(breast$Class, p = .2, list = FALSE) train[-labeled.index,cls] <- NA m <- WellSVMSSLR() %>% fit(Class ~ ., data = train) #Accuracy predict(m,test) %>% bind_cols(test) %>% metrics(truth = "Class", estimate = .pred_class)#> # A tibble: 2 x 3 #> .metric .estimator .estimate #> <chr> <chr> <dbl> #> 1 accuracy binary 0.706 #> 2 kap binary 0.0319