Nearly tight sample complexity bounds for learning mixtures of Gaussians via sample compression schemes

Open in new window