deep weight space

Graph Neural Networks for Learning Equivariant Representations of Neural Networks

Neural networks that process the parameters of other neural networks find applications in domains as diverse as classifying implicit neural representations, generating neural network weights, and predicting generalization errors. However, existing …