Learnable Similarity and Dissimilarity Guided Symmetric Non-Negative Matrix Factorization

Wenlong Lyu, Yuheng Jia · 2024

Read Learnable Similarity and Dissimilarity Guided Symmetric Non-Negative Matrix Factorization by Wenlong Lyu, Yuheng Jia (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04082.txt.

Description

Wenlong Lyu and Yuheng Jia improve a clustering method that splits data into like groups. Their twenty twenty-four paper works on symmetric nonnegative matrix factorization, a tool that groups items by a similarity map. The classic way builds that map from nearest neighbors, which can mislead.

Neighbors can belong to distinct clusters, and trust in them fades as the count grows. The new method gives each neighbor rank a learnable weight that states its reliability. This trims the search space sharply versus past methods. A paired dissimilarity map and a fresh orthogonality rule sharpen the split further. Tests on stability and geometry back the design.

Catalog record

CREATORWenlong Lyu, Yuheng Jia
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-04082

Browse the full library · About the Tower of Babel