By the end of this lecture
Four things you can do.
Soften a constraint
Turn a logical formula into a differentiable penalty, and say which t-norm you used.
Compute a semantic loss
Write −log P(α) for a small constraint and evaluate it on a distribution.
Read the family
Tell semantic loss, LTN, and DL2 apart by their semantics and their object of belief.
Name the limit
State why a loss is a prior, not a guarantee, and when to use an architecture or a verifier.