deep weight prior

A method that infuses prior knowledge into deep neural networks by learning a generative model of weights, helping to improve generalization in small-data regimes.