DS-MLR: Exploiting Double Separability for Scaling up Distributed Multinomial Logistic Regression (1604.04706v7)

Abstract: Scaling multinomial logistic regression to datasets with very large number of data points and classes is challenging. This is primarily because one needs to compute the log-partition function on every data point. This makes distributing the computation hard. In this paper, we present a distributed stochastic gradient descent based optimization method (DS-MLR) for scaling up multinomial logistic regression problems to massive scale datasets without hitting any storage constraints on the data and model parameters. Our algorithm exploits double-separability, an attractive property that allows us to achieve both data as well as model parallelism simultaneously. In addition, we introduce a non-blocking and asynchronous variant of our algorithm that avoids bulk-synchronization. We demonstrate the versatility of DS-MLR to various scenarios in data and model parallelism, through an extensive empirical study using several real-world datasets. In particular, we demonstrate the scalability of DS-MLR by solving an extreme multi-class classification problem on the Reddit dataset (159 GB data, 358 GB parameters) where, to the best of our knowledge, no other existing methods apply.