What is bias-variance tradeoff in deep learning?

Bias-variance tradeoff balances model simplicity and flexibility.

High bias leads to underfitting, high variance leads to overfitting. Deep learning tries to balance both via regularization and architecture design.