Self-Expanding Neural Networks

Rupert Mitchell, Robin Menzenbach, Kristian Kersting +1

The results of training a neural network are heavily dependent on the architecture chosen; and even a modification of only its size, however small, typically involves restarting th…