This paper explores the information bottleneck principle under the condition of perfect privacy, focusing on scenarios where the representation-rate constraint is active. The objective is to create a representation that maintains utility-relevant information while being statistically independent of a sensitive variable. This strict independence requirement adds a layer of complexity beyond the standard rate-relevance trade-off. To address this, the authors propose an alternating direction method of multipliers (ADMM)-based approach, demonstrating its global convergence and characterizing its rate using the Kurdyka-Lojasiewicz exponent, with extensions for inexact updates. AI
IMPACT This research contributes to theoretical understanding of privacy-preserving information representation, potentially influencing future AI model design.
RANK_REASON Academic paper on theoretical AI concepts. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX Code Finder for Papers
- CORE Recommender
- DagsHub
- Gotit.pub
- Hugging Face
- Influence Flower
- information bottleneck
- Kurdyka-Lojasiewicz exponent
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
