Under distributed differential privacy, the minimax error of federated PCA at the central server is the harmonic mean of the local clients' minimax errors, up to log factors.
(40) Meanwhile (see Appendix C.2), ∇ψ(Θ)∗ : Rp×p → Rp×r is given by ∇ψ(Θ)∗(M ) = ¯Θ⊥ ¯Θ⊤ ⊥(M + M ⊤) ¯ΘW −1/2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.ST 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Federated PCA and Estimation for Spiked Covariance Matrices: Optimal Rates and Efficient Algorithm
Under distributed differential privacy, the minimax error of federated PCA at the central server is the harmonic mean of the local clients' minimax errors, up to log factors.