Neural networks trained with SGD learn distributions of increasing complexity.

Maria Refinetti, Alessandro Ingrosso, Sebastian Goldt

Browse the full ICML paper archive.