Finding Manifolds With Bilinear Autoencoders

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dooms, Thomas, Gauderis, Ward
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908602102448128
author Dooms, Thomas
Gauderis, Ward
author_facet Dooms, Thomas
Gauderis, Ward
contents Sparse autoencoders are a standard tool for uncovering interpretable latent representations in neural networks. Yet, their interpretation depends on the inputs, making their isolated study incomplete. Polynomials offer a solution; they serve as algebraic primitives that can be analysed without reference to input and can describe structures ranging from linear concepts to complicated manifolds. This work uses bilinear autoencoders to efficiently decompose representations into quadratic polynomials. We discuss improvements that induce importance ordering, clustering, and activation sparsity. This is an initial step toward nonlinear yet analysable latents through their algebraic properties.
format Preprint
id arxiv_https___arxiv_org_abs_2510_16820
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Finding Manifolds With Bilinear Autoencoders
Dooms, Thomas
Gauderis, Ward
Machine Learning
Sparse autoencoders are a standard tool for uncovering interpretable latent representations in neural networks. Yet, their interpretation depends on the inputs, making their isolated study incomplete. Polynomials offer a solution; they serve as algebraic primitives that can be analysed without reference to input and can describe structures ranging from linear concepts to complicated manifolds. This work uses bilinear autoencoders to efficiently decompose representations into quadratic polynomials. We discuss improvements that induce importance ordering, clustering, and activation sparsity. This is an initial step toward nonlinear yet analysable latents through their algebraic properties.
title Finding Manifolds With Bilinear Autoencoders
topic Machine Learning
url https://arxiv.org/abs/2510.16820