Small feedback 💜
Hello 👋
@LenixC
great work, I love the multiple tabs where you can do model selection and hyperparameter tuning!
It would be super cool if you would add a small visualization of data in an intermediate place (maybe side by side with final result or next to dropdown itself?)
Also if you could elaborate more on what is going on in each tab in a simple basic language in the description it would be better for viewers.
Also shouldn't the strategies be radio button instead of checkbox?
Hey, @merve ! Yeah, I can definitely clarify what is going on for the reader. I originally was thinking of these spaces as sort of addendums to the already existing documentation and not as standalone interactive documentation.
So, the logic for using checkboxes for strategies is that the original post has a single graph that compares all of the learning strategies. I like seeing them all together, but it's a bit visually busy for me to actually just compare two or three. So, I like the checkboxes to solve that problem.
I can add visualizations, for sure! I think for moons and circles it is pretty obvious. However, for digits and iris, it's a bit less clear. What do you think about showing a handful of the handwritten numbers for the digits dataset? And for iris, I could either do PCA or select a 2 feature subset. PCA is more valuable for data representaiont, but I think sepal-length, sepal-height might be more intuitive for a viewer. What do you think?
Thanks!