Logistic Regression in Datastreams
Loading...
Date
2014-01
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Learning from data streams is a well researched task both in theory and practice. As remarked by Clarkson, Hazan and Woodruff, many classification problems cannot be very well solved in a streaming setting. For previous model assumptions, there exist simple, yet highly artificial lower bounds prohibiting space efficient one- pass algorithms. At the same time, several classification algorithms are often successfully used in practice. To overcome this gap, we give a model relaxing the constraints that previously made classification impossible from a theoretical point of view and under these model assumptions provide the first (1 + epsilon) -approximate algorithms for sketching the objective values of logistic regression and perceptron classifiers in data streams.