Saved in:
Bibliographic Details
Main Authors: Boufalis, Odysseas, Carrasco-Pollo, Jorge, Rosenthal, Joshua, Terres-Caballero, Eduardo, García-Castellanos, Alejandro
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.12601
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909905628168192
author Boufalis, Odysseas
Carrasco-Pollo, Jorge
Rosenthal, Joshua
Terres-Caballero, Eduardo
García-Castellanos, Alejandro
author_facet Boufalis, Odysseas
Carrasco-Pollo, Jorge
Rosenthal, Joshua
Terres-Caballero, Eduardo
García-Castellanos, Alejandro
contents Neural network parameterizations exhibit inherent symmetries that yield multiple equivalent minima within the loss landscape. Scale Graph Metanetworks (ScaleGMNs) explicitly leverage these symmetries by proposing an architecture equivariant to both permutation and parameter scaling transformations. Previous work by Ainsworth et al. (2023) addressed permutation symmetries through a computationally intensive combinatorial assignment problem, demonstrating that leveraging permutation symmetries alone can map networks into a shared loss basin. In this work, we extend their approach by also incorporating scaling symmetries, presenting an autoencoder framework utilizing ScaleGMNs as invariant encoders. Experimental results demonstrate that our method aligns Implicit Neural Representations (INRs) and Convolutional Neural Networks (CNNs) under both permutation and scaling symmetries without explicitly solving the assignment problem. This approach ensures that similar networks naturally converge within the same basin, facilitating model merging, i.e., smooth linear interpolation while avoiding regions of high loss. The code is publicly available on our GitHub repository.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12601
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Symmetry-Aware Graph Metanetwork Autoencoders: Model Merging through Parameter Canonicalization
Boufalis, Odysseas
Carrasco-Pollo, Jorge
Rosenthal, Joshua
Terres-Caballero, Eduardo
García-Castellanos, Alejandro
Machine Learning
Artificial Intelligence
Neural network parameterizations exhibit inherent symmetries that yield multiple equivalent minima within the loss landscape. Scale Graph Metanetworks (ScaleGMNs) explicitly leverage these symmetries by proposing an architecture equivariant to both permutation and parameter scaling transformations. Previous work by Ainsworth et al. (2023) addressed permutation symmetries through a computationally intensive combinatorial assignment problem, demonstrating that leveraging permutation symmetries alone can map networks into a shared loss basin. In this work, we extend their approach by also incorporating scaling symmetries, presenting an autoencoder framework utilizing ScaleGMNs as invariant encoders. Experimental results demonstrate that our method aligns Implicit Neural Representations (INRs) and Convolutional Neural Networks (CNNs) under both permutation and scaling symmetries without explicitly solving the assignment problem. This approach ensures that similar networks naturally converge within the same basin, facilitating model merging, i.e., smooth linear interpolation while avoiding regions of high loss. The code is publicly available on our GitHub repository.
title Symmetry-Aware Graph Metanetwork Autoencoders: Model Merging through Parameter Canonicalization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2511.12601