[BonoboBoner]

The first eight minutes were promising. Then a mathematical definiton of how neural networks work follows and I am already lost again when trying to learn about deep learning. It is like trying to learn Postgresql's new JSON features by starting to look at relational calculus first.

[argonaut]

Not sure why the other comment is being downvoted. You are looking at a talk given at the Simons Institute for the Theory of Computing. I'd be more worried if there wasn't any math.

If I wanted a newbie intro to operating systems, I would not be looking at talks at CS conferences like SOSP or OSDI, for example, even if they were "introductory."

[jph00]

Try http://course.fast.ai . We built it because we observed just the issue you have in most teaching materials for deep learning. We do cover all of the details eventually, but only when you're ready, and in a code first way.

[NicoJuicy]

Perhaps you'll like something at http://tagly.azurewebsites.net/Home/ByTag?Name=deep-learning

PS. It's a side project of mine, like HN with tags, needs improvement though :)

There's a lot about deep-learning subjects and some courses. If interested when launching with additional features : https://goo.gl/forms/dMx749vRcx30smT73

[tempw]

Each one has his/her pace and level, you need to find yours. There is a pletora of tutorial/courses available with a practical approach if you will.

If you know your math see The Deep Learning book for instance (which covers from math basics to current approaches).

Andrej K. also has a course focused on CV but covers the most important concepts of DL and CNN, RNN. DL/ML is nothing more than math afterall and a nontrivial subject, rather highly specialized so if you're in the rain is to get wet.