Skip to the content

Math 2.0

Resources / preprint

A Flow Matching Framework for Neural Representational Dissimilarity TX449S1VZV

URI	`m2://res/TX449S1VZV`
Kind	preprint
Creators	Zeyuan Ye, Xue-Xin Wei
Date	2026-09-25
Subjects	arxiv cs.AI, arxiv cs.IT, arxiv cs.LG, arxiv math.IT
Scope	border

Expressions

e1: version, preprint

Date	2026-09-25
Ids	arXiv 2609.31544

Items in this resource

The mentions of the inventory that name this resource. Each name links to the card of its inventory row.

Item	Kind	Where	Source
Averaging local affine coefficients gives linear Fisher information	statement	proposition A.8	arxiv
Fixed-norm translations give cosine distance	statement	proposition A.3	arxiv
Likelihood-trained translations give Euclidean, cosine, and correlation distances	statement	proposition B.1	arxiv
Condition-specific affine flows give Gaussian Jeffreys divergence	statement	proposition A.6	arxiv
Stimulus-dependent affine flows give Gaussian Fisher information	statement	proposition A.7	arxiv
Unconstrained flows give Jeffreys divergence and full Fisher information	statement	proposition A.9	arxiv
AYEDQQRR78	definition	definition	arxiv
Likelihood-trained affine flows give Mahalanobis and Gaussian Jeffreys divergence	statement	proposition B.2	arxiv
Similarity-constrained velocities generate similarity transformations	statement	proposition A.10	arxiv
Affine flows preserve Gaussianity	statement	lemma A.1	arxiv
Translation-constrained flow matching recovers squared Euclidean distance	statement	proposition A.2	arxiv
A shared centered affine flow gives Mahalanobis distance	statement	proposition A.5	arxiv
Likelihood-trained affine flows give Gaussian and linear Fisher information	statement	proposition B.3	arxiv
Centered fixed-norm translations give correlation distance	statement	proposition A.4	arxiv

Catalogue records

arxiv: 2609.31544 (work)