AI.info

Vicente Conde Mendes

Explore Vicente Conde Mendes on AI.info.

• A solvable high-dimensional model where nonlinear autoencoders learn structure invisible to PCA while test loss misaligns with generalization