Event Date
Speaker: Ting-Li Chen, Associate Research Fellow, Institute of Statistical Science, Academia Sinica (Taiwan) - visiting UC Davis 2024-2025
Title: "Towards Novel Categorical Embedding for Enhanced Data Representation"
Abstract: In the era of big data, categorical variables play a critical role across various domains, yet their analysis poses unique challenges due to the lack of a natural numerical representation. Traditional dimension reduction techniques, such as Principal Component Analysis (PCA), are designed for continuous data and cannot be directly applied to categorical variables without significant modifications.
In this talk, we introduce a novel approach to categorical data embedding that effectively maps categorical variables into a lower-dimensional Euclidean space. Our method extends the conceptual framework of PCA by optimizing the placement of categorical data to minimize predefined measurement errors. Through simulations and real-world applications, we demonstrate that our approach outperforms existing techniques such as Homogeneity Analysis (HOMALS) in preserving meaningful data structures for subsequent analysis. Finally, we discuss an extension of our framework to mixed-type data, enabling simultaneous dimension reduction for datasets containing both categorical and continuous variables.
Bio: Ting-Li Chen is an Associate Research Fellow at the Institute of Statistical Science, Academia Sinica, Taiwan. He received his Ph.D. from the Division of Applied Mathematics at Brown University. His research interests include Markov chain Monte Carlo, statistical learning, and high-dimensional data analysis.
Faculty web-page (links to Statistica Sinica): https://www.stat.sinica.edu.tw/eng/index.php?act=researcher_manager&code=view&member=7