https://arxiv.org/pdf/2110.11526#page=7&org=deepmind
Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth
Do Deep Convolutional Nets Really Need to be Deep and Convolutional?
Deep Information Propagation
Neural Networks, Manifolds, and Topology
High-performing neural network models of visual cortex benefit from high latent dimensionality
Scaling MLPs: A Tale of Inductive Bias