A nice thread using spline examples to explain the phenomenon in ML that the testing error can still decrease when the model complexity increases beyond the "sweet spot".
The Bias-Variance Trade-Off & "DOUBLE DESCENT" 馃УView quoted post on X
Remember the bias-variance trade-off? It says that models perform well for an "intermediate level of flexibility". You've seen the picture of the U-shape test error curve.
We try to hit the "sweet spot" of flexibility.
1/馃У