Thread to Discuss Books/MOOCs


#21

interested


(Rishubh Khurana) #22

Same here. Really interested.


(ecdrid) #23

(Sanyam Bhutani) #24

Guys do you suggest going through the Spring 2017 version of CS231N or the 2016 version?

Andrej Karparthy had taught the 2016 version.


(ecdrid) #25

Obviously 2016 version…
I had watched them till 10 lectures


(Sanyam Bhutani) #26

How is the Stanford NLP course?
Is it a wise idea to follow it?


#27

Despite how relatively little math I know, I find that my ability to do deep learning has been and continues to be bounded by ‘simple’ (simple is probably not the right word here), non-mathy things.

For instance, just right now I am using bcolz.ctables for the first time ever, a couple of days ago I learned a bit about Python generators, not too long ago I learned about ssh tunnels and use them to connect jupyter notebook running on my box…

What a reasonable person would do in such a scenario? Focus on the non-mathy things that seem to work! What does Radek do? Is attracted like a moth to a lightbulb to all sort of math related books despite all the evidence pointing to the contrary! :slight_smile: Guess what I asked Santa to bring me this Christmas :slight_smile: Ah the old wiring is not easy to overcome!

Well, thankfully not all is lost and through the guidance of Jeremy and Rachel and the through the copious amounts of time I sunk into things that didn’t bear fruit in the past, and also thx to not having too much time nowadays or at least as much time as I would like to give to DL, I tend to focus on the stuff that works ever so slightly better :slight_smile:


(helena s) #28

What does Radek do? Is attracted like a moth to a lightbulb to all sort of math related books despite all the evidence pointing to the contrary!

oh man, iso +1 on this. though i can reasonably hack my way around, the inability to appreciate all the math in the relevant DL papers bugs me to no end :cold_sweat:


(ecdrid) #29

@radek
You will be an amazing writer…


(Sanyam Bhutani) #30

I have to agree that the bottleneck in my performance is due to lack of ‘simple’ knowledge :slight_smile:
But then there is again the confusion, should I know this much theory for a little boost?
I found the CS231N-CNN Stanford course helpful. The assignments definitely help if the not the lectures (lectures are definitely theory intensive), assignments even though practical-are not very advanced topics. And I always find myself in the midst of the confusion.


(Zarak) #31

Here’s an upcoming book on building a bot that can play Go:


(Hongta) #32

new book to be available by François Chollet


#33

Could you please recommend a python MOOC that could serve as a prerequisite to this course?
Edit: I have just finished engineering degree, thus I have some programming experience. Also I did some basic programming in python already. I’m looking for a course that reviews the language, makes you think and is not terribly boring :wink:


(Joseph Catanzarite) #34

I found Andrew Ng’s deeplearning.ai specialization to be excellent preparation for fastai.