ICA can consistently bin similar sources together
Mathematics Exchange · 03.03.2026 · 17 (1) · 59-75
DOI: 10.33043/22467enb29 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics Exchange, online erschienen am 03.03.2026. Verfasst von Erin Krull und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Independent Component Analysis (ICA) is a blind-source separation method, meaning that it takes in a recording with multiple sensors and attempts to unmix it into the original sources. For example, suppose there are 4 people (sources) speaking in a room with 4 microphones (sensors), then ICA unmixes the recording from the 4 microphones to give tracks of the individual people called ICA components. ICA is currently used to decompose a variety of signals with many sensors, including fMRI and EEG data. However, its use in interpreting data with fewer sensors, such as the local field potential (LFP), is limited because of concerns about how it handles over-complete data (data with more sources than sensors). While there has been some success in enhancing ICA so that it can extract more sources than sensors, we focus on how ICA handles over-complete data. In this paper, we show that ICA consistently bins sources with similar spatial maps together when there are 3 sinusoidal sources and 2 sensors.
Zitieren
APA
Krull, E., Ollech, B., & Grabowski, K. (2026). ICA can consistently bin similar sources together. Mathematics Exchange, 17(1), 59-75. https://doi.org/10.33043/22467enb29
Vancouver
Krull E, Ollech B, Grabowski K. ICA can consistently bin similar sources together. Mathematics Exchange. 2026;17(1):59-75. doi:10.33043/22467enb29
BibTeX
@article{Krull2026,
author = {Krull, Erin and Ollech, Breanna and Grabowski, Kayley},
title = {ICA can consistently bin similar sources together},
journal = {Mathematics Exchange},
year = {2026},
volume = {17},
number = {1},
pages = {59--75},
doi = {10.33043/22467enb29}
}
EUVIMED