I have a data set that's ~ 150R X 2000C and was curious if an RBM is appropriate in situation with this type of imbalance. It's a microarray and I'm looking at a 0/1 classification problem. I'd be curious if there are any coded examples of using this especially in R/Python/Java.

  • $\begingroup$ I'm curious about it too. But can't you drop the number of features by using dimension reduction like PCA and then try to do some classification? $\endgroup$ – ThiS Jan 2 '13 at 10:29
  • $\begingroup$ What is the number of samples and features? Rows/columns differ from implementation to implementation. $\endgroup$ – bayerj Jan 2 '13 at 11:46
  • $\begingroup$ @bayerj: sorry, 150 samples / 2000 features. $\endgroup$ – screechOwl Jan 2 '13 at 16:51
  • $\begingroup$ In that case, my guess is that you have far too few samples. RBMs tend to work well if you have several thousands of samples with less features. What data are you looking at? Maybe I can recommend a similar method. $\endgroup$ – bayerj Jan 2 '13 at 16:56
  • $\begingroup$ @bayerj: so far I've looked at glmnet and bayesglm. Bayesglm seems to be noticeably better, but I'm not certain of the correct methodology for variable selection. $\endgroup$ – screechOwl Jan 2 '13 at 20:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.