search

suiji / Arborist

Scalable decision tree training and inference.
Other
82 stars 14 forks source link

set faults with some parameters #8

Closed topepo closed 8 years ago

topepo commented 8 years ago

Will investigate, repair and advise.

Thank you for reporting these.

mls

Here are two cases where Rborist fails:

library(caret)
library(Rborist)

dat <- twoClassSim(1000)

set.seed(1)
good1 <- Rborist(x = dat[, -ncol(dat)], y = dat$Class)

## segfault
set.seed(2)
bad1 <- Rborist(x = dat[, -ncol(dat)], y = dat$Class, nTree = 7)

set.seed(1)
good2 <- Rborist(x = dat[, -ncol(dat)], y = dat$Class, predFixed = 5)

## segfault on bagginig
ncol(dat)-1
set.seed(1)
bad2 <- Rborist(x = dat[, -ncol(dat)], y = dat$Class, predFixed = ncol(dat)-1)

Using Rborist_0.1-1 and Rcpp_0.12.4

suiji commented 8 years ago

These appear to be fixed by the latest changes on Github.

Closing, but feel free to reopen if needed.

Thanks again.