Modelling the Latent Semantics of Diffusion Sources in Information Cascade Prediction

Algorithm 1

Adversarial training framework in DSSF.

	Input: Training cascade sequences with diffusion sources
	;
	Number of epochs between training the discriminator ;
	Output: Optimized model parameters , , ;
(1)	Initialize , , ;
(2)	Calculate the cascade sequences number of each diffusion sources D
(3)	fordo
(4)	fordo
(5)	Sample a batch of training data B from S;
(6)	Divide into and evenly according D;
(7)	ifthen
(8)	Update the discriminator model: ;
(9)	else
(10)	Update the diffusion source semantics-enhanced encoder: ;
(11)	Update the cascade prediction layer: ;