Type change issue on page 433

[baifanhorst]

Running the codes on the top of page 433, I got this warning message: 'Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas.' I guess the reason is that the 'Cylinders' column contains integers while the other 4 numerical columns contain floating numbers. Although it is easy to convert the integers into floats, I am wondering whether we should treat this column as numeric. Maybe an ordered categorical (ordinal) feature is better.

[rasbt]

That's a good point and an interesting question. Coincidentally, I actually worked quite a bit on ordinal regression methods for deep neural networks :)

E.g.

• Deep Neural Networks for Rank-Consistent Ordinal Regression Based On Conditional Probabilities, https://arxiv.org/abs/2111.08851
• Rank consistent ordinal regression for neural networks with application to age estimation, https://arxiv.org/abs/1901.07884

I'd say in this case it's a count variable and one might make the case that the conversion to float is fine. On the other hand, the distance between the cylinders may not be well defined. E.g. the difference between 2 and 4 cylinders is maybe not the same as 6 and 8 cylinders. I'd say that the float conversion is probably fine though.