Wgan guide

[JamesAllingham]

Hey guys!

Sorry it's taken me so long to make this PR! Life has been getting in the way!

This is just a first draft and there is still a bunch I need to do, but I figured I should keep you all in the loop with what is going on here.

Here is what I still need to do:

1. A lot of proofreading, editing, and tweaking. I've done a rough first pass but I still need to read everything thoroughly to find spelling and grammar mistakes and change how things are worded.

2. Add a little more content. Mostly this will be in the form of motivations for some (or all) of the exercises.

3. Finish making the scribe notes. Currently, there are 7x 30-minute segments that need to be done. I've got volunteers for two of those and I'll find some time to do the rest. Once all of those are done I'll need to combine them, edit them, and link them to this guide.

4. Incorporate any important questions/answers/points from the notes into the main guide.

5. Incorporate Tali's demos somehow. I still need to talk to him to figure out how best to do this. He made them in Mathematica which most people won't have access too. Some options are to re-implement in another language, use Wolfram CDF, or make them non-interactive.

6. Tweak the GAN starter notebook to make it a bit easier for people. Maybe add a Pytorch one too (this could maybe come after we publish?). I have added links to Keras implementations of GAN, WGAN, and WGAN-GP for people who get stuck. But I still think it might be nice to have Pytorch because I suspect the implementations for WGAN and especially WGAN-GP will be easier.

And then, of course, I'm sure there will be plenty of feedback from you guys! :) Please let me know if you have any of that, any and all feedback is welcome! You are also welcome to wait until I've done some more proofing before you give feedback. I'll keep updating the PR as I make changes.

[suryabhupa]

Hey James! I'll spend some time going over the curriculum on the website in a bit more detail and provide feedback as it comes!

1. Sounds good, how do you want us to help you with this? I can request changes to the PR if that'd be helpful.

2. I think the motivations are an important part of the section so making sure it's well-written is important.

3. 👍 Thanks for doing this!

4. 👍

5. Yeah, that's a bit tricky. My first instinct would be to have them be in Colab, but it's up to you to decide what you mean by interactivity -- does a code sample with fill-in-the-blanks fulfill this, or would you rather have cool interactive js diagrams? Up to you (though Colabs are always a safe bet).

6. Feel free to add the PyTorch one later, because I think it's probably good as it stands. We can first prioritize getting the Mathematica --> Colabs, and the scribe notes, and then adding a PyTorch implementation.

I'll start editing the PR over the next few days.

[JamesAllingham]

1. That sounds great. I'm going to find some time to do some editing in the next 24 hours, so after that, any additional changes you can recommend would be very helpful - before then you'll probably just be finding many small and silly mistakes.

2. I agree. I'll also find some time to do the first draft of those in the next 24 hours.

3. Cool, will get back to you about this ASAP :)

4. Okay, that seems like a reasonable way to prioritize this.

Thanks!

[JamesAllingham]

Okay, I've incorporated most of (all?) of your feedback and I've also added the notes from our sessions! I think we can take this live now! 🎈

There are some things left to do:

1. Add Pytorch solution notebooks. As discussed before, we can launch without these and add them later since we do already have Keras solutions.
2. Add Tali's stuff. Having looked at the recordings and spoken to him about this, I think we probably overvalued this. It won't end up being an interactive thing, but rather some cool, but static, visualizations in the notes for that week. I don't think it is worth delaying for this.
3. Go through the notes and generalize them in some places so that they are more useful for folks who weren't in the sessions with us. I think the notes are still useful for the most part, just that they could be slightly improved in some cases, so let's not wait on this either IMO. 🎉

Let me know what you think, but I am happy and excited to launch this!

Oh, one other thing that needs to happen before launching is to change the date in the name of the markdown file as well as in the header at the top 😸

[suryabhupa]

This all looks quite good! I'm quite pleased with it now -- fantastic work, James! :) I'll go ahead and do the merge now.

[suryabhupa]

@JamesAllingham Can you just do final checks and make sure it renders properly, as well as finalize all acknowledgements, and edit the file name? Let's close this after that.