Antonio Vergari

Discrete causal inference with hidden confounders is now tractable and scalable, matching neural models' accuracy with millisecond inference.

Faithful concept bottleneck explanations are impossible without including all concepts, but probabilistic modeling and group lasso enable compact, faithful explanations.

Uncertainty-quantified neural surrogates can drastically enhance Bayesian inference efficiency for complex inverse problems, outperforming traditional methods in high-dimensional settings.

Predicting unseen links in knowledge graphs can often hinge on just a single half-link, revealing critical insights into model generalization.

Even when your variational approximation is wrong, symmetries in the target distribution can guarantee you still get the mean right.