An Efficient Alternating Newton Method for Learning Factorization Machines

Recently, factorization machines (FM) have emerged as a powerful model in many applications. In this work, we study the training of FM with the logistic loss for binary classification, which is a non-linear extension of the linear model with the logistic loss (i.e., logistic regression). For the training of large-scale logistic regression, Newton methods have been shown to be an effective approach, but it is difficult to apply such methods to FM because of the non-convexity. We consider a modification of FM that is multi-block convex and propose an alternating minimization algorithm based on Newton methods. Some novel optimization techniques are introduced to reduce the running time. Our experiments demonstrate that the proposed algorithm is more efficient than stochastic gradient algorithms. The parallelism of our method is also investigated for the acceleration in multi-threading environments.