PARAFAC2 is a powerful method for analyzing multi-modal data consisting of irregular frontal slices. In this work, we propose POPLAR method that imposes graph Laplacians constraints induced by the similarity symmetric tensor as auxiliary information to force decomposition factors to behave similarly and the method is developed using AO-ADMM for 3-way PARAFAC2 tensor decomposition. To the best of our knowledge, POPLAR is the first approach to incorporate graph Laplacians constraints using auxiliary information. We extensively evaluate \method's performance in comparison to state-of-the-art approaches across synthetic and real datasets and POPLAR clearly exhibits better performance with respect to the Fitness (better 3-8%), and F1 score (better 5-20%) among the state-of-the-art factorization method. Furthermore, the running time for the method is comparable to the state-of-art method.

PARAFAC2 Decomposition; Tensor