active_subspaces

New method enhances AI uncertainty quantification for neural networks

Researchers have developed a novel method for scalable uncertainty quantification in neural networks, crucial for high-stakes AI applications. Their approach, termed generalized Laplace active subspaces, constructs a lo…