----------------------------------------------------------------------------------------------------

         Department of Systems Engineering and Engineering Management

                             The Chinese University of Hong Kong

----------------------------------------------------------------------------------------------------



Date: 4:30pm - 5:30pm on 26 September (Friday)

Venue: ERB 513, The Chinese University of Hong Kong

Title: Generative Choice Models for Subset Selection

Speaker: Yu Yang, Department of Data Science, City University of Hong

Kong



Abstract:

Choice modeling, which explores how users select from a finite set of

items, is vital for operational decision-making. While it's common for

users to purchase multiple items in a single transaction, much of the

existing literature has focused solely on the case of users choosing one

item at a time. A more realistic choice model must account for the

correlations among items. In this talk, I will present our recent work

on learning generative choice models for subset selection. I will first

highlight the challenges associated with learning a generative choice

model from data by showcasing key statistics from a real customer order

dataset. Following this, I will introduce an intuitive and effective

metric for evaluating generative choice models, demonstrating that

employing biased models can actually be beneficial. Motivated by this

insight, we propose two generative choice models based on representation

learning and deep sequence generative models, along with algorithms to

learn their parameters from data. Extensive experiments on various

real-world datasets show that our choice models effectively capture the

underlying subset distribution and enhance downstream operations

management applications. Finally, I will discuss learning a subset

choice model through the lens of data compression, proposing a novel

tree-based set compression method that captures conditional independence

among items while maintaining efficiency in model complexity.







Biography:

Yu Yang is an Associate Professor in the Department of Data Science at

City University of Hong Kong. His research interests focus on the

algorithmic aspects of data mining and data science, particularly in

developing effective algorithms for analyzing data with combinatorial

structures (such as graphs, sets, and sequences) and for data-driven

operations management. He also applies his work to various applications,

including social marketing, supply chain management, traffic pattern

detection, and smart medical wearables. His research has been published

in premier venues such as SIGMOD, VLDB, ICDE, ICML, NeurIPS, ICLR,

AISTATS, KDD, TKDE, and TKDD. He currently serves as an associate editor

for ACM Transactions on Knowledge Discovery from Data. Yu obtained his

Ph.D. in Computing Science from Simon Fraser University in 2019, and

previously earned his B.E. from Hefei University of Technology in 2010

and his M.E. from the University of Science and Technology of China in

2013, both in Computer Science.