% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{Albers:1028773,
author = {Albers, Jasper and Kurth, Anno and Gutzen, Robin and
Morales-Gregorio, Aitor and Denker, Michael and Grün, Sonja
and van Albada, Sacha and Diesmann, Markus},
title = {{Q}uantifying shared structure between real matrices of
arbitrary shape},
reportid = {FZJ-2024-04819},
year = {2024},
abstract = {Assessing the similarity of matrices is valuable for
analyzing the extent to which data sets exhibit common
features in tasks such as data clustering, dimensionality
reduction, pattern recognition, group comparisons, and graph
analysis. Methods proposed for comparing vectors, such as
the cosine similarity, can be readily generalized to
matrices. However, these approaches usually neglect the
inherent two-dimensional structure of matrices. Existing
methods that take this structure into account are only
well-defined on square, symmetric, positive- definite
matrices, limiting the range of applicability. Here, we
propose Singular Angle Similarity (SAS), a measure for
evaluating the structural similarity between two arbitrary,
real matrices of the same shape based on singular value
decomposition. By taking the two-dimensional structure of
matrices explicitly into account, SAS is able to capture
structural features that cannot be identified by traditional
methods such as Euclidean distance or the cosine
similarity.After introducing and characterizing the measure,
we apply SAS to two neuroscientific use cases: adjacency
matrices of probabilistic network connectivity, and state
evolution matrices representing neural brain activity. We
demonstrate that SAS can distinguish between network models
based on their adjacency matrices. Furthermore, SAS captures
differences in high-dimensional responses to different
stimuli in MUAe data from macaque V1, which can be related
to the underlying response properties of the neurons.
Thereby, SAS allows for a quantification of closeness of
related response patterns in a network of neurons. We
conclude that SAS is a suitable measure for quantifying the
shared structure of matrices with arbitrary shape in
neuroscience and beyond.},
month = {Jun},
date = {2024-06-25},
organization = {FENS Forum, Wien (Austria), 25 Jun
2024 - 29 Jun 2024},
subtyp = {After Call},
cin = {IAS-6 / INM-10},
cid = {I:(DE-Juel1)IAS-6-20130828 / I:(DE-Juel1)INM-10-20170113},
pnm = {5231 - Neuroscientific Foundations (POF4-523) / BMBF
03ZU1106CB - NeuroSys: Algorithm-Hardware Co-Design (Projekt
C) - B (BMBF-03ZU1106CB) / DFG project 313856816 - SPP 2041:
Computational Connectomics (313856816) / EBRAINS 2.0 -
EBRAINS 2.0: A Research Infrastructure to Advance
Neuroscience and Brain Health (101147319) / JL SMHB - Joint
Lab Supercomputing and Modeling for the Human Brain (JL
SMHB-2021-2027)},
pid = {G:(DE-HGF)POF4-5231 / G:(DE-Juel1)BMBF-03ZU1106CB /
G:(GEPRIS)313856816 / G:(EU-Grant)101147319 / G:(DE-Juel1)JL
SMHB-2021-2027},
typ = {PUB:(DE-HGF)24},
url = {https://juser.fz-juelich.de/record/1028773},
}