Welcome to Part 1 (v2)

(Jeremy Howard) #1

Use this category for discussions of Practical Deep Learning for Coders (2018), Part 1. Be sure to check the FAQ before posting, and read about how to ask for help.

Here are threads to help you with specific issues:

Finally, here’s the introduction thread where you can get to know the other students and introduce yourself.

Unofficial release of part 1 v2
Interested in joining a learning cohort for part 1 in early January?
Unofficial release of part 1 v2
Interested in joining a learning cohort for part 1 in early January?
Setting of learning rate
Lesson 1 - ImportError with Theano
(Soumya) #2

Simple Stats I could infer from the excel sheet that you provided

We have 325 remote students registered so far.

Top 10 time-zones:

UTC+5:30 85
UTC-5 48
UTC+1 28
UTC+8 25
UTC-8 21
UTC+3 17
UTC+2 14
UTC-4 13
UTC-7 11
UTC-3 8

Top 10 Locations:

Bangalore,India 26
India 13
Chennai,India 6
London,UK 5
Australia/Perth 5
Mumbai,India 5
Delhi,India 4
Raleigh,NC 4
Singapore 4
Melbourne,Australia 4

Looks like a lot of students from India and specifically from Bangalore!

PS: I have done very basic data-wrangling on this. I too am from Bangalore, India :grinning:

(Olivier Ma) #3

Great chance for you guys to have some local meetups! I have to go hundreds of miles to find some one :frowning:

(Sanyam Bhutani) #4

That really looks like an opportunity to have some meet ups in Bangalore at some point of time. :slight_smile:

(Sudhanshu Garg) #5

Cool stats. Bangalore people, let’s connect on Slack to plan a meetup: https://join.slack.com/t/blr-fast-ai/shared_invite/enQtMjU5NTY5MzI2MzcxLTRhZWJlYTc0ZjI4MzBkNjA0NTZmYzlmN2RlOTFmZjA3MjQ3Njc1MjNiODY3MDYxMWQwN2YwYTFmMTI4ODVmOWI

(Sanyam Bhutani) #6

Imo a WhatsApp group would be a better idea, I don’t think a better way to discuss related code than the forums, why create another slack ‘mini-community’? We could always have our little thread on here.

(vikram iyer) #7

Whatsapp and Slack will be too noisy (hard to maintain and regain info, though easy for short notices), some other way we can collaborate on projects and take this further?

(Vikram Kalabi) #8

Google groups is a viable option

(Jeremy Howard) #9

Feel free to create a “Bangalore” thread here on this category, if that’s useful :slight_smile:

(Aditya) #10

Is there any curated list of the datasets Jeremy will be using in this course?
I will download them beforehand…

(Jeremy Howard) #11

Yes. See /datasets/ in your crestle.com instance.

(naveen manwani) #12

@jeremy ,today i started working with paperspace ,as soon as i gave my card details ,1$ was debited from my credit card .so i need to enquire that will they credit back that 1 $ just like AWS and google cloud service do.i know it just 1$ but currently i’m unemployed so every penny matter for me.and though i have given the card details but still i haven’t received mail from paperspace regarding activation of P5000 machine on my paperspace account.could you please provide give me some insights about my query.

thanks in advance

(Stephen Lizcano) #13

They usually debit 1$ to ensure the validity of the card, not overlimit, etc. I would assume it would be put back within a few days.

(naveen manwani) #14

thank you for your kind reply


@jeremy Just a general question - is it possible to cover reinforcement learning this time.

(Jeremy Howard) #16

No we won’t cover that.

(Mudit Verma) #17

can you provide some resources on it ?

(Satya Prateek) #18

There are some good RL courses freely available. David Silver’s RL course is the classic one (https://www.youtube.com/watch?v=2pWv7GOvuf0&list=PL7-jPKtc4r78-wCZcQn5IqyuWhBZ8fOxT)

There’s a course on deep reinforcement learning (http://rll.berkeley.edu/deeprlcourse/) that looks great but I haven’t gone through it yet.

Finally there’s a nice github project (https://github.com/dennybritz/reinforcement-learning) that has implementations for a lot of the popular RL algs.

I remember reading a lot of good blog posts/tutorials but I can’t find them now. If I do I’ll add them in.

(sergii makarevych) #19

Nice sources, thanks @satya


Google GPU offering seems to be very slow compared to Amazon’s ec2 for the same price.