We are excited to announce that our paper “Permuton-induced Chinese restaurant process” has been accepted to NeurIPS2021 as a poster presentation.
This paper proposes a multi-dimensional extension of the Chinese restaurant process (CRP) and applies it to Bayesian nonparametric relational data analysis. Our proposed model can be used as a unified stochastic process to represent various classes of rectangular partitioning. This model can be regarded as a multi-dimensional extension of CRP paired with the block-breaking process (BBP), which has been recently proposed as a multi-dimensional extension of SBP. While BBP always has an infinite number of redundant intermediate variables, our model can be composed of varying size intermediate variables in a data-driven manner depending on the size and quality of the observation data. Quantitative experiments demonstrate that our model can improve the prediction performance in relational data analysis by reducing the local optima and slow mixing problems compared with the conventional BNP models, because the local transitions in Markov chain Monte Carlo inference are more flexible.
We have released a Matlab implementation at GitHub. A python implementation will be coming soon.
Pre-proceedings can be accessible from the following link:
Permuton-induced Chinese Restaurant Process
Permuton-induced Chinese Restaurant Process Part of Advances in Neural Information Processing Systems 34 pre-proceedings (NeurIPS 2021) Bibtek download is not available in the pre-proceeding Authors Masahiro Nakano, Yasuhiro Fujiwara, Akisato Kimura, Takeshi Yamada, naonori ueda Abstract This paper proposes the permuton-induced Chinese restaurant process (PCRP), a stochastic process on rectangular partitioning of a matrix.