Fig. 9From: Distributed stochastic gradient descent for link prediction in signed social networksSensitivity to the rank estimate r. The loss function is squared hinge, the number of computational threads is M=10, and the regularization parameter is λ=0.2Back to article page