<ul><li>We study the task of learning latent-variable models.</li><li>We develop a general efficient algorithm for implicit moment tensor computation.</li><li>The algorithm enables poly-time learning algorithms for mixtures of linear regressions, mixtures of spherical Gaussians, and positive linear combinations of non-linear activations.</li><li>The complexity of the algorithm depends on the desired error and the target class of functions.</li></ul>

Implicit High-Order Moment Tensor Estimation and Learning Latent Variable Models

Discover more