Add GIFs explaining unsupervised learning algorithms?

[rhiever]

This post has some nice GIFs explaining some unsupervised learning algorithms. Each time I've taught this workshop, I wish I had GIFs like these to explain e.g. k-means. What do you think about adding them?

Pinging the author, @dashee87, to get his opinion.

[amueller]

Not sure. I often prefer static images. This way I can pick my own speed in the "animation".

If you have forward and backward buttons to step through the algorithm, that might be the ideal?

[amueller]

I like the dbscan and mean-shift ones, though. The K-means was a bit hard to follow for me. I don't have a dbscan algorithm plot because I felt you might need too many plots to show the progress. Maybe that's not true, I'm not sure.

[dashee87]

Hey guys,

The GIFs were always intended to act as educational tools, so you certainly have my permission to use them as you wish. I'm happy to provide the python files that generated each set of images. That would allow you, in some respects, to control the speed. Or I could add some annotations to the GIFs (e.g. for kmeans, include some text saying 'Input Data', 'Initialisation', etc.). But I get what you're saying. Static images allow you to spend more time on particular aspects of the algorithm (just click to the next slide when you've finished).

[rhiever]

@dashee87, would you be able to provide the static images that compose the GIFs? Then we could do as @amueller suggested.

[dashee87]

That's not a problem. I'll upload a folder to my blogScripts repository. It'll take a few days, as I want to add some annotations to the kmeans gif.

[dashee87]

I've uploaded the static images here. Note that I've added a slightly altered version of the 'good' kmeans gif, as well as the original. Some filenames are a little on the lazy side (1.png, 2.png, etc), so apologies if it's initially a little confusing. Let me know if you need anything else.