Fetching the paper…

Projected BNNs: Avoiding weight-space pathologies by learning latent representations of neural network weights · Around