Hacker News new | past | comments | ask | show | jobs | submit login
Organized Resources for Deep Learning Researchers and Developers (github.com/astorfi)
153 points by irsina on Aug 13, 2018 | hide | past | favorite | 22 comments

I am an undergrad trying to break into AI and machine learning research. Hopefully this will be useful. Does anyone know any other resources that will be helpful? So far I've just been reading AI: A Modern Approach, by Russel and Norvig.

It's hard to recommend something if we don't really know your level of knowledge so far.

I wholeheartedly recommend the fast.ai [0] course. It provides a lot of instantly applicable code, coupled with very good explanations which you can try out on novel problems later. It's focused on "learning by doing", and not "learning by reading" which fits my style really well.

That said, it doesn't dissuade the watcher from reading later, it's just not recommended to start out with.

[0]: course.fast.ai

I think this class/site is mentioned on every relevant HN for very good reason: it's actually that good.

I dislike learning from video (upping playback speed helps), I dislike the coding style of the library and the notebooks (nonlinear notebook execution especially), and I still think this is the best available class on anything deep-learning related, and it's only getting better. The top-down, practice-before-theory approach is excellent, but they still get into the theory, often in a much more intuitive and better motivated way than you get elsewhere. Also tons of little breadcrumbs dropped throughout lessons and in the forum to dig deeper for those inclined to.

If you go this route, make sure to follow the suggestion of re-implementing each lesson, from scratch, without referring back to the original notebook. It's a little too easy to not do that and miss out on the lessons you learn from struggling through the actual code.

So far I have taken one introductory class on AI in general, but it did not cover machine learning. I took one class on machine learning, but I only grasped the basics of several algorithms from the class. These classes, and the textbook I've partially completed, are the extent of my knowledge.

Thank you for that suggestion.

Take as much probability and linear algebra as you can conveniently do – as much for the intuition as for the symbol-manipulation mechanics – and don't underrate the importance of domain expertise in any problem you get interested in!

I would highly recommend going through François chollet's Deep Learning with Python[1] book. The technical concepts are explained very well and since you have gone through the Modern AI book you won't have an issue understanding them. It's a very hands-on book and by the time you finish it you will be able to use Deep Neural Nets to solve many problems.

I would also recommend going through the scikit-learn documentation. Some of the tutorials/examples there are pretty good.

At the end of the day, it all comes down to your personal learning style. For me the thing that worked was to go through the above mentioned steps and then find a problem I was interested in and try to solve it using my newly found skills. That way you will discover new tools and methods.

Finally, the Deep Learning [2] book is also very good but I would not recommend it to a beginner. It's better to use it when you have a basic understanding of Machine Learning and you want to gain a deeper understanding of the concepts.


[2]: https://www.deeplearningbook.org/

I would pair [1] with Hands On Machine Learning with Scikit-Learn and Tensorflow by Aurélien Géron (I own both). It gives an excellent overview of machine learning including non-deep stuff (plus the ins and outs of scikit-learn and tensorflow).

If you haven't already studied Linear Algebra, and want to get a headstart on that, check out the "Coding The Matrix" book/videos from Brown.




Also, see the Gilbert Strang video series on Linear Algebra:


and the amazing 3blue1brown "Essence of Linear Algebra" series:


I thought both of these two courses on coursera were quite good:



First one is a bit older school, but takes you through all the fundamentals and actually explains a lot of the math involved. It also gets you thinking a lot more about how to solve problems from a Linear Algebra standpoint and the types of problems machine learning is good for tackling.

Second one is a much more modern day set of courses specifically focused on Deep Learning techniques and problem solving.

I thought both were great. First one is free as well...

"Pattern Recognition and Machine Learning" by Christopher Bishop.


I like "Machine Learning: A Probabilistic Perspective" by Kevin Murphy more, but this really is a matter of taste; both are excellent.

Sutton & Barto, there’s a new edition due in October

worth noting this is reinforcement-learning specific; fascinating field, getting a lot of press the last few years, but best considered an important and distinct field that just happens to intersect with ML/DL. I'd suggest understanding DL on simpler problems (well-understood CV/NLP problems) before wading into using it in reinforcement learning.


We took "Show HN" out of the title. A list of resources doesn't count as something people can try out, as the rules describe at https://news.ycombinator.com/showhn.html.

I made this Quora post, which might be useful too: http://qr.ae/TUTB1t

This is very neat! I hope I can make use of this :)

I have added a couple of issues to the repo that have been my historic "blockers" towards exploring deep learning: data sets and computation images (docker images, AMIs, etc).

I made this Quora post you might find it useful, http://qr.ae/TUTB1t

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact