Just tried this out, very cool. Making data wrangling fast is at least as important as model selection.
The way "deep learning as a service" is shaping up, there are 3 camps: vertical focused, platform as black box, and platform for power users.
"Vertical focused" would include companies like Clarifai and Enlitic focused on image tagging and medical imaging respectively.
"Platform as a black box" would be companies like Metamind, and apparently Labellio (just upload and hit "train").
"Platform for power users" would be Ersatz Labs (full control of model type, parameter selection, ensembling, focus on supporting newer-but-trickier model types like sequence-to-sequence RNNs).
Jury is out on which approach is the right one. For the record, my company makes Ersatz, launched in January 2013.
The title reads like something out of Silicon Valley (the TV show). "We're changing the world by providing location-aware in-app database technology for everyone."
The way "deep learning as a service" is shaping up, there are 3 camps: vertical focused, platform as black box, and platform for power users.
"Vertical focused" would include companies like Clarifai and Enlitic focused on image tagging and medical imaging respectively.
"Platform as a black box" would be companies like Metamind, and apparently Labellio (just upload and hit "train").
"Platform for power users" would be Ersatz Labs (full control of model type, parameter selection, ensembling, focus on supporting newer-but-trickier model types like sequence-to-sequence RNNs).
Jury is out on which approach is the right one. For the record, my company makes Ersatz, launched in January 2013.