Fitting tabular data using smooth curve fits

If you're fitting a curve or surface to data and want to ever use gradient-based optimizers or Newton solvers, make sure to do it in a differentiable way.

0:00 - Intro
1:15 - Using a piecewise linear fit is generally bad
2:25 - A smooth and continuous curve fit is advantageous
5:09 - Further notes and recommendations
6:26 - Conclusion