BPTF和BGCP区别在哪里？

[sequelOoO]

BPTF和BGCP区别在哪里？

[xinychen]

Thank you for this good question! BPTF is used to the data in recommender systems, and the data has the dimensions of user, product, and time. So as can be seen, BPTF has a smoothing regularization for time dimension. BGCP is a more general model that takes a basic form of tensor factorization and a fully Bayesian treatment. This model can be used to any tensor data. A further discussion of BPTF is our another work:

• https://arxiv.org/pdf/1910.06366.pdf

That is Bayesian temporal tensor factorization (BTTF). Despite of missing data imputation, BTTF can be applied to multidimensional time series forecasting.

Any further comment would be appreciated, thank you!