Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Deep Learning for Coders – Launching Deep Learning Part 2

fast.ai

Post reply on HN
Print view View original

11–20 of 94 posts

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#11

Post by pzh »

BrianMingus wrote: post #4

Latently (SUS17) also provides a more self-directed path to learning deep learning focused exclusively on implementing research papers and conducting original research: https://github.com/Latently/DeepLearningCertificate

That seems interesting, but there are so many papers and little indication of which ones are more important (or which ones to implement first). I realize this is for advanced learners, but some guidelines, or at least a section pointing to survey papers would be really helpful as a starting point.

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#12

Post by daedalus13 »

jph00 wrote: post #7
daedalus13 wrote: post #6

jph00, I found the first course hard to follow because of some broken links and poorly organized content. One link that was necessary kept taking me to a password protected page. This is about a month ago. It would be good if someone could revisit part.1 and make those minor editorial fixes if they haven't already done so. I might be being too precious about my time, but I also found the first video about your teachi…

The move from platform.ai to files.fast.ai could have been communicated better - sorry this impacted you. (We tried to highlight everywhere we could, but we can't change the video itself on YouTube unfortunately.) We're redoing the whole of part 1 starting in October so this problem will be fully resolved then. Until then, follow the links on course.fast.ai or the forums, rather than what you see in the part 1 videos…

thanks. I'd also like to suggest publishing a docker image as well for course software. Like you I like to also read / experiment when travelling and an offline workflow can often be handy.

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#13

Post by jph00 »

daedalus13 wrote: post #12
jph00 wrote: post #7

Earlier quoted context omitted.

The move from platform.ai to files.fast.ai could have been communicated better - sorry this impacted you. (We tried to highlight everywhere we could, but we can't change the video itself on YouTube unfortunately.) We're redoing the whole of part 1 starting in October so this problem will be fully resolved then. Until then, follow the links on course.fast.ai or the forums, rather than what you see in the part 1 videos…

thanks. I'd also like to suggest publishing a docker image as well for course software. Like you I like to also read / experiment when travelling and an offline workflow can often be handy.

There's a docker image available : https://github.com/deeprig/fastai-course-1 . This is from the same excellent person who created http://crestle.com , which has all the notebooks and data files for the course pre-installed.

(You can also install everything locally using the setup scripts available on the course github repo.)

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#14

Post by drieddust »

jph00 wrote: post #10
jph00 wrote: post #7

Earlier quoted context omitted.

The move from platform.ai to files.fast.ai could have been communicated better - sorry this impacted you. (We tried to highlight everywhere we could, but we can't change the video itself on YouTube unfortunately.) We're redoing the whole of part 1 starting in October so this problem will be fully resolved then. Until then, follow the links on course.fast.ai or the forums, rather than what you see in the part 1 videos…

Oh and about the teaching philosophy video - until we posted that we had quite a few students express confusion about the top-down approach. After posting it, we've received a lot of positive feedback about it. I understand it's not helpful or interesting to everyone, but overall it seems to have been a successful addition for most.

I am one of those guys who likes to understand everything before beginning to do anything meaningful. Needless to say I mostly fail.

Your lecture on top down course philosophy actually helped me change my perspective. It's a welcome positive change so thank you.

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#15

Post by jacquesm »

throwaway12017 wrote: post #5

What is the goal of these trainings? To get a taste so you understand the conversation? There is a lot more to data science than neural networks, and I'm skeptical that teaching one family of models will create a set of implementers that don't compare and contrast solutions.

Practical applications rather than theoretical knowledge. In other words, if you're a programmer to put more tools in your toolbox.

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#16

Post by jacquesm »

jph00 wrote: post #13
daedalus13 wrote: post #12

Earlier quoted context omitted.

thanks. I'd also like to suggest publishing a docker image as well for course software. Like you I like to also read / experiment when travelling and an offline workflow can often be handy.

There's a docker image available : https://github.com/deeprig/fastai-course-1 . This is from the same excellent person who created http://crestle.com , which has all the notebooks and data files for the course pre-installed. (You can also install everything locally using the setup scripts available on the course github repo.)

It might be worth it to do an updated version of course #1 at some point, many of the examples will not work now if you try to follow along with the video because libraries (for example Keras) have been changed or upgraded.

(I know you have plenty to do, so really no pressure, this is simply a nice to have for newcomers that 'tuned in late'.)

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#17

Post by jph00 »

jacquesm wrote: post #16
jph00 wrote: post #13

Earlier quoted context omitted.

There's a docker image available : https://github.com/deeprig/fastai-course-1 . This is from the same excellent person who created http://crestle.com , which has all the notebooks and data files for the course pre-installed. (You can also install everything locally using the setup scripts available on the course github repo.)

It might be worth it to do an updated version of course #1 at some point, many of the examples will not work now if you try to follow along with the video because libraries (for example Keras) have been changed or upgraded. (I know you have plenty to do, so really no pressure, this is simply a nice to have for newcomers that 'tuned in late'.)

Yup we're starting it in October! :)

(Although if folks follow the course and use the AMI or setup scripts provided, all the correct versions will be installed automatically still today.)

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#18

Post by jph00 »

phunge wrote: post #3

Highly recommended! The first course was the first thing I came across that helped me contextualize the DL field into something that might be relevant for my work. It's a great way to get your hands dirty. One point of comparison is Cam Davidson Pilon's Bayesian Methods for Hackers, they have a similar vibe: practical applied advice from a field that tilts towards the academic...

That's a very flattering comparison! :) Such a great book - here's a link for those interested in bayesian methods: http://nbviewer.jupyter.org/github/CamDavidsonPilon/Probabil...

In fact that book inspired me to create a spreadsheet that implements MCMC in order to make it easy to understand and visualize - we're planning to start an "Introduction to Machine Learning" course in a couple of months where I hope to show off the result of this...

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#19

Post by throwaway12017 »

jacquesm wrote: post #15
throwaway12017 wrote: post #5

What is the goal of these trainings? To get a taste so you understand the conversation? There is a lot more to data science than neural networks, and I'm skeptical that teaching one family of models will create a set of implementers that don't compare and contrast solutions.

Practical applications rather than theoretical knowledge. In other words, if you're a programmer to put more tools in your toolbox.

Sound fair. I'll have to watch the videos to get a better opinion.

Re: Deep Learning for Coders – Launching Deep Learning Part 2

#20

Post by BrianMingus »

pzh wrote: post #11
BrianMingus wrote: post #4

Latently (SUS17) also provides a more self-directed path to learning deep learning focused exclusively on implementing research papers and conducting original research: https://github.com/Latently/DeepLearningCertificate

That seems interesting, but there are so many papers and little indication of which ones are more important (or which ones to implement first). I realize this is for advanced learners, but some guidelines, or at least a section pointing to survey papers would be really helpful as a starting point.

We have a bibliography - it is not yet organized very well but we are working on that: https://paperpile.com/shared/UhfbVO
Post reply on HN