Aishwarya Venkataramanan

We introduce Distance-informed Neural Processes (DNP), a Neural Process variant with bi-Lipschitz regularization that preserves input geometry. DNP yields better-calibrated uncertainty & stronger OOD detection across regression + classification.