We study the problem of decomposing a low-rank matrix into a factor with binary entries, either from {+1,-1} or from {0,1}, and an unconstrained factor. Such binary component decompositions are appropriate
for applications where the latent factor re
ects an exclusive choice (e.g. "on" and "off" in electrical
engineering; "connected" or "disconnected" in graph theory; "yes" and "no" in survey data; "like" and "dislike" in collaborative ?filtering; or "active" and "inactive" in genomics). Our research answers fundamental questions about the existence and uniqueness of these decompositions. It also leads to
tractable factorization algorithms that succeed under a mild deterministic condition.
This is joint work with Joel Tropp (Caltech).
Sprache der Kurzfassung:
Englisch
Vortragstyp:
Eingeladener Vortrag an anderen Institutionen
Vortragsdatum:
15.04.2021
Vortragsort:
Österreich
Details zum Vortragsort:
Johann Radon Institute for Computational and Applied Mathematics Austrian Academy of Sciences
Forschungs-