Generic orbit recovery from invariants of very low degree

Dan Edidin, Josh Katz
{"title":"Generic orbit recovery from invariants of very low degree","authors":"Dan Edidin, Josh Katz","doi":"arxiv-2408.09599","DOIUrl":null,"url":null,"abstract":"Motivated by the multi-reference alignment (MRA) problem and questions in\nequivariant neural networks we study the problem of recovering the generic\norbit in a representation of a finite group from invariant tensors of degree at\nmost three. We explore the similarities and differences between the descriptive\npower of low degree polynomial and unitary invariant tensors and provide\nevidence that in many cases of interest they have similar descriptive power. In\nparticular we prove that for the regular representation of a finite group,\npolynomial invariants of degree at most three separate generic orbits answering\na question posed in \\cite{bandeira2017estimation}. This complements a\npreviously known result for unitary invariants~\\cite{smach2008generalized}. We\nalso investigate these questions for subregular representations of finite\ngroups and prove that for the defining representation of the dihedral group,\npolynomial invariants of degree at most three separate generic orbits. This\nanswers a question posed in~\\cite{bendory2022dihedral} and it implies that the\nsample complexity of the corresponding MRA problem is $\\sim \\sigma^6$. On the\nother hand we also show that for the groups $D_n$ and $A_4$ generic orbits in\nthe {\\em complete multiplicity-free} representation cannot be separated by\ninvariants of degree at most three.","PeriodicalId":501475,"journal":{"name":"arXiv - MATH - Commutative Algebra","volume":"57 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - MATH - Commutative Algebra","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2408.09599","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Motivated by the multi-reference alignment (MRA) problem and questions in equivariant neural networks we study the problem of recovering the generic orbit in a representation of a finite group from invariant tensors of degree at most three. We explore the similarities and differences between the descriptive power of low degree polynomial and unitary invariant tensors and provide evidence that in many cases of interest they have similar descriptive power. In particular we prove that for the regular representation of a finite group, polynomial invariants of degree at most three separate generic orbits answering a question posed in \cite{bandeira2017estimation}. This complements a previously known result for unitary invariants~\cite{smach2008generalized}. We also investigate these questions for subregular representations of finite groups and prove that for the defining representation of the dihedral group, polynomial invariants of degree at most three separate generic orbits. This answers a question posed in~\cite{bendory2022dihedral} and it implies that the sample complexity of the corresponding MRA problem is $\sim \sigma^6$. On the other hand we also show that for the groups $D_n$ and $A_4$ generic orbits in the {\em complete multiplicity-free} representation cannot be separated by invariants of degree at most three.
从极低度不变量中恢复通用轨道
受多参考对齐(MRA)问题和不等变神经网络问题的启发,我们研究了从最多为三度的不变量张量中恢复有限群表示中的泛位的问题。我们探讨了低度多项式不变张量和单元不变张量描述力的异同,并提供证据证明在许多感兴趣的情况下,它们具有相似的描述力。特别是,我们证明了对于有限群的正则表达,度数最多为三个的多项式不变式会分离出一般轨道,这回答了在《bandeira2017estimation》中提出的一个问题。这补充了之前已知的单元不变式结果~(cite{smach2008 generalized}。我们还针对有限群的次规则表示研究了这些问题,并证明对于二面体群的定义表示,度数为至多三个的多项式不变式会分离出一般轨道。这回答了在~cite{bendory2022dihedral}中提出的一个问题,它意味着相应的 MRA 问题的样本复杂度是 $\sim \sigma^6$。另一方面,我们还证明了对于$D_n$和$A_4$群,{em complete multiplicity-free}表示中的泛轨道不能被最多三度的变量分开。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信