A spotlight paper to be presented in NeurIPS2020

Baxter permutation process

We are excited to announce that our paper “Baxter permutation process” has been accepted to NeurIPS2020 as a spotlight presentation.

In this paper, a Bayesian nonparametric model for Baxter permutations (BPs), termed BP process (BPP) is proposed and applied to relational data analysis. The BPs are a well-studied class of permutations, and it has been demonstrated that there is one-to-one correspondence between BPs and several interesting objects including floorplan partitioning, which constitutes a subset of rectangular partitioning. Accordingly, the BPP can be used as a floorplan partitioning model. We combine the BPP with a multi-dimensional extension of the stick-breaking process called the block-breaking process to fill the gap between floorplan partitioning and rectangular partitioning, and obtain a stochastic process on arbitrary rectangular partitionings. Compared with conventional Bayesian nonparametric models for arbitrary rectangular partitionings, the proposed model is simpler and has a high affinity with Bayesian inference.

You can find a pre-proceedings paper at https://proceedings.neurips.cc/paper/2020/hash/6271faadeedd7626d661856b7a004e27-Abstract.html and a MATLAB implementation at https://github.com/nttcslab/baxter-permutation-process .