R 3.2.3 new H2o Package (2015) cannot achieve the same results from old package -


i using h2o r package (2014 version) perform deep learning task using textual data. did research in 2015 , obtained promising results using deep learning method (function - h2o.deeplearning; e.g. fscore , recall achieve >0.9). found original r code doesn't work (due change of h2o package in nov 2015) , revised code. however, when tried run same deep learning model (same setting), not achieve outperfom results anymore!! please, wish know if h2o has changed internal modeling settings since revision of h2o package? wish reproduce old results new package... please kindly help.

h2o deep learning (2.0 , 3.0) not reproducible default -- can change setting reproducible = true, slow things down quite bit, reproducibility requires code run on single core. therefore variability due randomness in algorithm alone, rather upgrade of h2o 2.0 3.0.

if want use h2o classic (2.0), old code still work, is. might try running first see if can track down source of variability. there nothing wrong using h2o classic finish project started while ago.

implementation details h2o 3.0 deep learning available in deep learning booklet.

there more information on has changed in h2o dl between h2o 2.0 , 3.0 here.


Comments

Popular posts from this blog

javascript - jQuery: Add class depending on URL in the best way -

caching - How to check if a url path exists in the service worker cache -

Redirect to a HTTPS version using .htaccess -