[edit]
Algorithms for learning a mixture of linear classifiers
Proceedings of The 33rd International Conference on Algorithmic Learning Theory, PMLR 167:205-226, 2022.
Abstract
Linear classifiers are a basic model in supervised learning. We study the problem of learning a mixture of linear classifiers over Gaussian marginals. Despite significant interest in this problem, including in the context of neural networks, basic questions like efficient learnability and identifiability of the model remained open. In this paper, we design algorithms for recovering the parameters of the mixture of $k$ linear classifiers. We obtain two algorithms which both have polynomial dependence on the ambient dimension $n$, and incur an exponential dependence either on the number of the components $k$ or a natural separation parameter $\Delta>0$. These algorithmic results in particular settle the identifiability question under provably minimal assumptions.