Hyper-Representations: Learning from Populations of Neural Networks

TL;DR

This paper introduces hyper-representations, a self-supervised approach to learn general representations of neural network weights, enabling better understanding, sampling, and transfer of models across architectures and tasks.

Contribution

It proposes hyper-representations as a novel method to learn task-agnostic, meaningful structures in neural network weights, facilitating interpretability and transfer learning.

Findings

Neural networks occupy meaningful structures in weight space.

Hyper-representations can predict model performance and training state.

They enable sampling of models with targeted properties.

Abstract

This thesis addresses the challenge of understanding Neural Networks through the lens of their most fundamental component: the weights, which encapsulate the learned information and determine the model behavior. At the core of this thesis is a fundamental question: Can we learn general, task-agnostic representations from populations of Neural Network models? The key contribution of this thesis to answer that question are hyper-representations, a self-supervised method to learn representations of NN weights. Work in this thesis finds that trained NN models indeed occupy meaningful structures in the weight space, that can be learned and used. Through extensive experiments, this thesis demonstrates that hyper-representations uncover model properties, such as their performance, state of training, or hyperparameters. Moreover, the identification of regions with specific properties in hyper-representation space allows to sample and generate model weights with targeted properties. This thesis demonstrates applications for fine-tuning, and transfer learning to great success. Lastly, it presents methods that allow hyper-representations to generalize beyond model sizes, architectures, and tasks. The practical implications of that are profound, as it opens the door to foundation models of Neural Networks, which aggregate and instantiate their knowledge across models and architectures. Ultimately, this thesis contributes to the deeper understanding of Neural Networks by investigating structures in their weights which leads to more interpretable, efficient, and adaptable models. By laying the groundwork for representation learning of NN weights, this research demonstrates the potential to change the way Neural Networks are developed, analyzed, and used.