[Unofficial, Virtual] SF/Bay Area Study Group

Possible Ch2 related topics we could cover:

  1. What kind of tabular data is deep learning NOT particularly good at?
  2. How might an e-commerce company like Amazon tackle the downsides of deep learning based recommendation systems such as: recommending books by the same author, or items someone has already purchased?
  3. Map the steps of the Drivetrain approach to a Kaggle contest (for ex: Titanic, Amazon Rainforests, or your choice)
  4. Take a past work (or side) project you’ve done that had a data component and imagine a deep learning model was applied. What would be an example of “domain shift” for that project?
  5. For a project you’re interested in applying deep learning to, a) Describe how the 3 stages of deployment processes (manual process, limited scope, gradual expansion) would work. b) Pretend you’ve completed that future project already, and it went really, really well? Describe what’s happened as a result.
  6. Share the link to an image recognition model you created using data you curated that is deployed on the web.
  7. Share the link to a deep learning blog post you wrote.
1 Like

If you haven’t already, please fill out the anonymous survey for the Tuesday study groups: https://docs.google.com/forms/d/e/1FAIpQLSdsk_Uk-8gu1XlQ2DKVdQnqsR-AIcwppFwpBjBSUqm_Qi0eJA/viewform?usp=sf_link

Looking for both people who regularly attend, and those who have not had a chance to attend.

Hi all,

For our biweekly online presentation meetings (next one scheduled for Thursday, July 16 at 6pm) - @tyoc213 and I (@butchland) would like to present our work-in-progress project on getting fastai2 running on a TPU using Pytorch-XLA.

If you would like to join, the meeting details are here

Our project code is here

Best regards,
Butch

2 Likes

@butchland looking forward to it!

1 Like

Updating the meeting link
and the schedule to 7PM San Francisco time.

2 Likes

Hello, for anyone who is watching is interested in the second half of the notebooks, please fill out this form: https://forums.fast.ai/t/part-2-notebook-review/75269

Results are in! It seems Mondays are the best, though we will be having two times. Starting at 12PM-3PM, and 6PM-9PM our regular time, though on Monday.

@butchland @tyoc213 Can you both make next Monday 6-9PM? Also feel free to message anyone else you know is interested.

@pinaki

Yes I’m available for the 6-9 pm slot.
Thanks!

@butchland @pinaki @tyoc213 Here is the new google calendar link: https://calendar.google.com/calendar/embed?src=s59fhs8iql3iuttdnjp7t3ad8g%40group.calendar.google.com&ctz=America%2FLos_Angeles

2 Likes

Hi @marii,
Can you post the meeting link?
I’m not sure if my calendar invite has the right one.

Thanks,
Butch

Hi @marii
which notebook will we be covering today ?

I previously suggested CNN chapter (13 I think) but we can discuss alternatives if you want.

Best regards
Butch

Hey Molly,

What is the study group looking at these days? Is the group doing the later lessons?

Thanks,
Daniel

Yes, we are doing the later lessons. Next notebook is 13_convolutions.ipynb.

We are meeting on Mondays now, the link to the google calendar is a few messages above.

FYI I’ve updated the post at the top of this thread with Monday evening details + link to your calendar. I believe you can also edit that post @marii.

@steef I think I signed up to present something. I have a few things I can prepare, but was there something I specifically signed up for to present on Thursday?

I have you on the agenda for optimizers. Does that work? Honestly anything that you’d be happy to present on will be more than welcome! Thanks Molly! @marii

1 Like

Sure! Sounds great. I just didn’t remember if I had signed up or not. I already have it ready, so no problems here.

1 Like

Due to members participating in hackathon our study group was delayed on week. Would have updated earlier, but could not access the forum. See you next week for the notebook group, or this Thursday for steef’s group!

Hi Everyone,

I’m just starting the new version of the Deep Learning (Part 1 - 2020) class and I’m guessing that there will be others that are just starting it as well, since my understanding is that it was just released the the general web in the past couple of weeks.

Would it make sense to start a new Pacific Time Zone study group (on Zoom) for those who are at a similar stage of study? If anyone is interested, or if anyone thinks that this would be a bad idea for some reason, please let me know!

Quick background on me… I have taken a couple of college classes in computer science and I have a few years of professional software development experience but it’s out of date (Visual Basic). I went through Part 1 of the previous class but I didn’t “go deep” into the extra work and exercises to really master the material. I’m planning to go much deeper this time around.

1 Like