A Fast Active Learning Algorithm for Link Classification
published: Aug. 6, 2013, recorded: April 2013, views: 2509
Report a problem or upload filesIf you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.
We present a very efficient active learning algorithm for link classification in signed networks. Our algorithm is motivated by a stochastic model in which edge labels are obtained through perturbations of an initial sign assignment consistent with a two-clustering of the nodes. We provide a theoretical analysis within this model, showing that we can achieve an optimal (to within a constant factor) number of mistakes on any graph G = (V;E) such that |E| = \Omega(|V|^3/2) by querying O(|V|^3/2) edge labels. More generally, we show an algorithm that achieves optimality to within a factor of O(k) by querying at most order of |V| + (|V|/k)^3/2 edge labels. The running time of this algorithm is at most of order |E| + |V| log |V|.
Download slides: machine_zappella_learning_algorithm_01.pdf (255.8 KB)
Link this pageWould you like to put a link to this lecture on your homepage?
Go ahead! Copy the HTML snippet !