For well-separated Gaussian mixtures, over-parameterized gradient EM with n=Omega(m log m) components converges globally to the ground truth, the first such result beyond m=2.
Extreme ratio between spectral and frobenius norms of nonnegative tensors
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Global Convergence of Gradient EM for Over-Parameterized Gaussian Mixtures
For well-separated Gaussian mixtures, over-parameterized gradient EM with n=Omega(m log m) components converges globally to the ground truth, the first such result beyond m=2.