A perturbed ADMM algorithm with convergence-rate guarantees is developed for finite-alphabet information bottleneck problems that impose exact independence from a sensitive variable.
Information extraction under privacy constraints,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IT 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Information Bottleneck under Perfect Privacy
A perturbed ADMM algorithm with convergence-rate guarantees is developed for finite-alphabet information bottleneck problems that impose exact independence from a sensitive variable.