The Communication Complexity of Distributed Estimation
Analysis
This article from Apple ML delves into the communication complexity of distributed estimation, a problem where two parties, Alice and Bob, aim to estimate the expected value of a bounded function based on their respective probability distributions. The core challenge lies in minimizing the communication overhead required to achieve a desired accuracy level (additive error ε). The research highlights the relevance of this problem across various domains, including sketching, databases, and machine learning. The focus is on understanding how communication scales with the problem's parameters, suggesting an investigation into the efficiency of different communication protocols and their limitations.
Key Takeaways
- •The research focuses on the communication complexity of distributed estimation.
- •The goal is to estimate the expected value of a function with additive error.
- •The problem has applications in sketching, databases, and machine learning.
“Their goal is to estimate Ex∼p,y∼q[f(x,y)] to within additive error ε for a bounded function f, known to both parties.”