July 24, 2013


I’ve been working on the Kaggle Amazon Employee Access competition. I got a late start and am in a rush to try out a new model. Unfortunately, the current R implementation is slow.

I left it running overnight and woke to find it almost done. Just when my excitement was peaking, I hit a bug and the routine stopped. I didn’t write the code in a way that checkpointed the progress so all six hours of computation is lost. Biscuits!

Notes to self:

  • Try harder to make faster models.
  • Write code that checkpoints progress.
  • Don’t wait until the final days to start a Kaggle competition.

Tags:  Kaggle , R