Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

neural network weights

Neural network weights are adjustable numerical parameters that determine the strength and direction of the connections between artificial neurons across layers in a neural network. During the training process, these values are iteratively modified by optimization algorithms, such as gradient descent paired with backpropagation, to minimize the error between the network predictions and the target outcomes. By scaling the incoming signals passed between connected nodes, weights effectively encode the learned features, patterns, and operational knowledge of the model. Together with bias terms, weights form the primary internal parameters that define how a trained machine learning model transforms raw input data into final predictions.

1 item

Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights

Hyper-Representations as Generative Models: Sampling Unseen Neural Network Weights

Konstantin Schürholt, Boris Knyazev, Xavier Giró-i-Nieto, Damian Borth

OrganizationsAIML Lab, School of Computer ScienceInstitut de Robòtica i Informàtica Industrial, CSIC-UPCSamsung SAILUniversitat Politècnica de CatalunyaUniversity of St. Gallen

Why you should read this

Introduces a generative approach using layer-wise loss normalization to sample diverse, high-performing neural network weights directly from model zoos for effective initialization, ensembling, and transfer learning without requiring underlying training data.

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.

Added

2026-09-26