cy-dev / sparsesvm

Solution Paths of Sparse Linear Support Vector Machine with Lasso or ELastic-Net Regularization

sparseSVM

Solution Paths of Sparse Linear Support Vector Machine with Lasso or ELastic-Net Regularization

To install:

• the released version from CRAN: install.packages("sparseSVM")
• the latest version (require devtools): install_github("CY-dev/sparseSVM")

To report bugs：

• send email to: Congrui Yi [email protected] and Yaohui Zeng [email protected]

sparsesvm's Issues

Parameter `dfmax` not working properly.

set.seed(2)
X <- matrix(0, 200, 1000)
X[] <- rnorm(length(X))
y <- sample(0:1, size = nrow(X), replace = TRUE)

test <- sparseSVM::sparseSVM(X, y, dfmax = 20)
print(colSums(test$weights != 0))

Each time, it stops before evaluating the 100 values of lambda, but always a little too late.
It should have stopped at lambda = 0.1229 but continues evaluating at 6 other values of lambda.