Neural network weight size matches shortest algorithm length

Original: Why does deep learning generalize? What does weight decay really do? Can algorithmic information theory address these questions?

Source: x.com ↗