We are excited to announce that our paper “Nonparametric relational models with superrectangulation” has been accepted to AISTATS2022 as an oral presentation. Among 1685 submissions, 492 papers (29%) were accepted and only 44 (top 9% of accepted papers) of them have been selected as orals.
This paper addresses the question, ''What is the smallest object that contains all rectangular partitions with or fewer blocks?'' and shows its application to relational data analysis using a new strategy we call super Bayes as an alternative to Bayesian nonparametric methods. We propose a strategy to combine an extremely redundant rectangular partition as a deterministic (non-probabilistic) object. Specifically, we introduce a special kind of rectangular partitioning, which we call superrectangulation, that contains all possible rectangular partitions. Delightfully, this strategy completely eliminates the difficult task of searching around for random rectangular partitions, since the superrectangulation is deterministically fixed in inference. Experiments on predictive performance in relational data analysis show that the super Bayesian model provides a more stable analysis than the existing Bayesian models, which are less likely to be trapped in bad local optima.
You can find the proceeding paper at https://proceedings.mlr.press/v151/nakano22a.html .
Nonparametric Relational Models with Superrectangulation
Masahiro Nakano, Ryo Nishikimi, Yasuhiro Fujiwara, Akisato Kimura, Takeshi Yamada, Naonori Ueda Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:8921-8937, 2022. Abstract This paper addresses the question, "What is the smallest object that contains all rectangular partitions with n or fewer blocks?"