mariofilho

I uploaded the image to Gemini 2.0 Flash Thinking 01 21 and asked:

“ Here is a steps vs bmi plot. What do you notice?”

Part of the answer:

“Monkey Shape: The most striking feature of this plot is that the data points are arranged to form the shape of a monkey. This is not a typical scatter plot where you'd expect to see trends or correlations between variables in a statistical sense. Instead, it appears to be a creative visualization where data points are placed to create an image.”

Gemini 2.0 Pro without thinking didn’t see the monkey

God forbid I put a link to my blog when sharing a useful tool for free!

This is something I got curious about too. It's very likely the answer is no, but I would like to test it at some point.

It will very likely fail to predict new data. The model can approximate the interactions over the range it saw on training, but we need to show it larger ranges if we want to predict for a bigger grid.

Hi, thanks for your comment:

I opened an issue on Github. This didn't seem obvious to many people that read the article, so it's nice that now we can keep this in mind while using the model:

https://github.com/dmlc/xgboost/issues/4069

Hi, thanks for your comment.

You are correct that this model will not generalize to numbers outside this range.

My goal was just to have a reference when this questions comes up in a discussion about creating feature interactions that are about differences, multiplications, etc.

So I wanted to show that, yes, the model is capable of approximating it. Of course, we would need a sample that covers the necessary range to generalize.

This is a good topic for a future article.