Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights
This work addresses the challenge of knowledge aggregation from model zoos to create new models, offering potential applications in model inspection and neural architecture search, but it is incremental as it builds on existing hyper-representation methods.
The authors tackled the problem of generating new neural network weights from a model zoo by extending hyper-representations for generative use, achieving diverse and high-performing models that outperform strong baselines on downstream tasks like initialization and transfer learning.
Learning representations of neural network weights given a model zoo is an emerging and challenging area with many potential applications from model inspection, to neural architecture search or knowledge distillation. Recently, an autoencoder trained on a model zoo was able to learn a hyper-representation, which captures intrinsic and extrinsic properties of the models in the zoo. In this work, we extend hyper-representations for generative use to sample new model weights. We propose layer-wise loss normalization which we demonstrate is key to generate high-performing models and several sampling methods based on the topology of hyper-representations. The models generated using our methods are diverse, performant and capable to outperform strong baselines as evaluated on several downstream tasks: initialization, ensemble sampling and transfer learning. Our results indicate the potential of knowledge aggregation from model zoos to new models via hyper-representations thereby paving the avenue for novel research directions.