Logistic Regression in Datastreams

Learning from data streams is a well researched task both in theory and practice. As remarked by Clarkson, Hazan and Woodru [12], many classi cation problems cannot be very well solved in a streaming setting. For previous model assumptions, there exist simple, yet highly arti cial lower bounds prohibiting space e cient onepass algorithms. At the same time, several classi cation algorithms are often successfully used in practice. To overcome this gap, we give a model relaxing the constraints that previously made classi cation impossible from a theoretical point of view and under these model assumptions provide the rst (1 + )-approximate algorithms for sketching the objective values of logistic regression and perceptron classi ers in data streams.