Cornell University Library [https://arxiv.org/] has a lot of publicly available papers. You need to find the domain that you are interested in and start looking. Usually, papers describe relative works and state of the art algorithms. So you can start with one algorithm and go deep.

Books: https://medium.com/iambot/tagged/books

Also, you can join communities.

https://deeplearningstudies.slack.com

https://www.facebook.com/groups/1892696574296664/

The latest papers are posted to http://arxiv.org, but a more curated approach is very helpful.

So take a look at http://www.gitxiv.com and http://www.arxiv-sanity.com