The tradeoff between a model being too simple and too fitted to examples.
Memorize every old math answer, and one new question ruins your day. Learn only “math is numbers,” and good luck.
This tradeoff helps set how bendy a model should be. You meet it when choosing models, tuning settings, and fighting overfitting.
Regularization
Regularization helps stop overfitting and balance variance.
ERM
ERM can push variance up when it only chases low training error.
SLT
This is one of the classic tradeoffs in SLT.
Inductive Bias
A model’s prior assumptions directly change its bias and variance.