Efficient Online Inference for Nonparametric Mixture Models

Rylan Schaeffer, Blake Bordelon, Mikail Khona, Weiwei Pan, Ila Rani Fiete

Uncertainty in Artificial Intelligence Accepted

July 2021

Bayesian Nonparametrics Online Learning Mixture Models Clustering

PDF Poster Slides

Abstract

We derive efficient online inference algorithms for nonparametric mixture models, enabling streaming clustering that grows in complexity as needed by the data.

Summary

Efficient online inference algorithms for nonparametric mixture models.

Summary

This work develops efficient online inference algorithms for Dirichlet Process mixture models, enabling clustering on streaming data where the number of clusters grows as needed.

Key Contributions:

1. Derivation of recursive forms of the Chinese Restaurant Process for streaming data
2. Efficient algorithms that match or exceed batch inference quality
3. Significantly reduced computational requirements compared to offline methods