Definition

It is seems to be the case that over-parameterising the model does not cause overfitting. center

Where, Hypothesis Space

Note: The Interpolation threshold is the point where the model has exactly enough parameters to perfectly fit all training data (zero training error).