Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Under the hood, all the existing (as of 2.0) ML pipelines models immediately convert the Dataframe to an RDD and cache it, so there is a one iteration slowdown for using the Dataframe models versus an equivalently implemented RDD model. None of them try to get a speedup by directly operating on the Tungsten datastructure, but it is an obvious next step.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: