3 ms·
IDK about those products but Google's AutoML stuff is probably the best of breed here. A few research groups have published AutoML-like stuff on github too. Li
by grizzles 8y ago
IDK about those products but Google's AutoML stuff is probably the best of breed here. A few research groups have published AutoML-like stuff on github too.
Like anything it depends on your use case. Sometimes even a tiny bit of feature engineering can have a huge impact on model fidelity.
- eyeball 8y agoThe google one looked like it was focused only on a few use cases (language / vision). I've played around with some of the open source stuff (tpot, autosklearn, h20's automl). The commercial products seem to do more automated feature engineering. H2o's automl is pretty easy to use with their graphical "flow" notebook interface. Was able to build a reasonable model with no coding aside from prepping the data going in.