Teasing Apart Architecture and Initial Weights as Sources of Inductive Bias in Neural Networks

Open in new window