A random-order streaming algorithm approximates the top eigenvector with near-linear memory whenever the spectral gap is constant, and a lower bound shows the heavy-row parameter is unavoidable.
First efficient convergence for streaming k- PCA : a global, gap-free, and near-optimal rate
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DS 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Approximating the Top Eigenvector in Random Order Streams
A random-order streaming algorithm approximates the top eigenvector with near-linear memory whenever the spectral gap is constant, and a lower bound shows the heavy-row parameter is unavoidable.