The Distinguished Lecture Series of School of Computing and Data Science, host distinguished scholars around the world who will share their expertise and insights in the areas of computer science, data science, artificial intelligence, and statistics.

Our 3rd Distinguished Lecture, titled “Poisson Hierarchical Indian Buffet Processes – with indications for Microbiome Models”, will be presented by Professor Lancelot F. JAMES, a Chair Professor at HKUST Business School.

Please find the event details below:

Date: 30 April, 2025 (Wednesday)

Time: 4:30pm – 5:30pm (Reception starts at 4:00pm)

Venue: CB-A, G/F, Chow Yei Ching Building, Main Campus, The University of Hong Kong
Medium: English

Speaker:

Professor Lancelot F. JAMES
Director, MSc In Business Analytics
Chair Professor
Information Systems, Business Statistics, and Operations Management
HKUST Business School

Abstract:

We describe Poisson Hierarchical Indian Buffet Processes, designed for complex random sparse count species sampling models that facilitate information sharing across and within groups in various contexts. This model accommodates a potentially infinite number of species (taxa) and unknown parameters, allowing us to learn as more data is gathered within a Bayesian machine learning framework. We focus on the challenging context of microbiome analysis and related ecological species sampling models to address existing gaps in modelling capabilities.

Our model offers a generative process for these phenomena, providing flexible sparse multivariate count models that account for overdispersion while also modelling latent OTU counts. We present tractable methods for sampling and posterior analysis in this complex setting, introducing novel parameters that reflect species abundance as well as alpha and beta diversity. We also provide indications for novel approaches to the formidable problem of unseen entities in future samples.

Biography:

Professor Lancelot F. JAMES, along with co-author Hemant Ishwaran, has significantly contributed to popularizing key concepts such as stick-breaking priors, the generalised Chinese restaurant process, and Pitman-Yor processes, which have been vital in Bayesian nonparametric statistics and related machine learning applications since the early 2000s. He also developed the Poisson Partition Calculus, central to the analysis of various Bayesian nonparametric models, including latent feature models. In recognition of these and other contributions, he was elected as a Fellow of the Institute of Mathematical Statistics in 2008. He was born on the island of Jamaica.

All are welcome to attend.