[polkapolka]

If you spend one year on applying deep learning, you can train a net on a 100 different data sets. That's where the intuition comes from. You'll debug a lot. People with zero experience with deep learning have ended in the top 10 for Kaggle computer vision competitions. Applied ML does not require a PhD in Computer Science. It is not like you are debugging Cuda as was the case a few years back. Even then, a PhD in statistics is unlikely to help with that. Really, one year of Python experience and you are good to go.

How rare is it to find a new problem domain these days? Most if not all problems tackled with deep learning have already been done in one form or the other. Deep learning is not yet much of an academic discipline, but accessible to anyone willing to put in the hours of study and practice.

[nostrademons]

"How rare is it to find a new problem domain these days? Most if not all problems tackled with deep learning have already been done in one form or the other."

Curious if anyone's done this: deep learning on DOM trees, to recognize portions of HTML pages. This has elements in common with both CNNs (where you want surrounding context - parents, ancestors, siblings, and descendants - to influence the recognition of a particular DOM element) and with RNNs (because elements may have an unbounded number of children, and potentially a recurring substructure), but I don't know offhand how you would combine them to solve the problem.

Anyone working on this? A quick literature search turned up nothing.

[bolololo1]

Unsupervised learning to recognize patterns?

Seems like a really interesting idea to work on.