Uncovering the Spectral Bias in Diagonal State Space Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Solozabal, Ruben, Bojkovic, Velibor, AlQuabeh, Hilal, Inui, Kentaro, Takáč, Martin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918131916603392
author Solozabal, Ruben
Bojkovic, Velibor
AlQuabeh, Hilal
Inui, Kentaro
Takáč, Martin
author_facet Solozabal, Ruben
Bojkovic, Velibor
AlQuabeh, Hilal
Inui, Kentaro
Takáč, Martin
contents Current methods for initializing state space models (SSMs) parameters mainly rely on the \textit{HiPPO framework}, which is based on an online approximation of orthogonal polynomials. Recently, diagonal alternatives have shown to reach a similar level of performance while being significantly more efficient due to the simplification in the kernel computation. However, the \textit{HiPPO framework} does not explicitly study the role of its diagonal variants. In this paper, we take a further step to investigate the role of diagonal SSM initialization schemes from the frequency perspective. Our work seeks to systematically understand how to parameterize these models and uncover the learning biases inherent in such diagonal state-space models. Based on our observations, we propose a diagonal initialization on the discrete Fourier domain \textit{S4D-DFouT}. The insights in the role of pole placing in the initialization enable us to further scale them and achieve state-of-the-art results on the Long Range Arena benchmark, allowing us to train from scratch on very large datasets as PathX-256.
format Preprint
id arxiv_https___arxiv_org_abs_2508_20441
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Uncovering the Spectral Bias in Diagonal State Space Models
Solozabal, Ruben
Bojkovic, Velibor
AlQuabeh, Hilal
Inui, Kentaro
Takáč, Martin
Machine Learning
Artificial Intelligence
Current methods for initializing state space models (SSMs) parameters mainly rely on the \textit{HiPPO framework}, which is based on an online approximation of orthogonal polynomials. Recently, diagonal alternatives have shown to reach a similar level of performance while being significantly more efficient due to the simplification in the kernel computation. However, the \textit{HiPPO framework} does not explicitly study the role of its diagonal variants. In this paper, we take a further step to investigate the role of diagonal SSM initialization schemes from the frequency perspective. Our work seeks to systematically understand how to parameterize these models and uncover the learning biases inherent in such diagonal state-space models. Based on our observations, we propose a diagonal initialization on the discrete Fourier domain \textit{S4D-DFouT}. The insights in the role of pole placing in the initialization enable us to further scale them and achieve state-of-the-art results on the Long Range Arena benchmark, allowing us to train from scratch on very large datasets as PathX-256.
title Uncovering the Spectral Bias in Diagonal State Space Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2508.20441