Email: allen29@ucla.edu / ykei@ucsc.edu
Google Scholar

I am a Visiting Assistant Professor in the Department of Statistics at University of California, Santa Cruz (UCSC). I obtained a Ph.D. in Statistics and Data Science at University of California, Los Angeles (UCLA) in June 2024, under the supervision of Prof. Oscar Hernan Madrid Padilla.

My research focuses on developing realistic models and the associated inference and learning algorithms for data to allow useful abstraction, generation, and detection. I believe that abstracting meaningful representations from multi-modal data can help devise powerful models that generate valid knowledge in a real-world setting, while detecting meaningful structures in data under certain criteria.

My research has also been advised by Prof. Mark S. Handcock and Prof. Ying Nian Wu at UCLA, Prof. Robert B. Lund, Prof. Rebecca Killick, Prof. James D. Wilson during my current position at UCSC, and many other amazing collaborators.

My research interests are

• Generative Models
• Representation Learning
• Multi-modal Learning
• Graph Inference
• Anomaly and Cluster Detection
• Empirical Bayes Methodology
• Optimization

Previously, I have interned at Amazon and Cisco. I received a B.S. in Statistics and a B.A. in Economics at UCLA.

Publications

Published papers

1. A collaborative paper has been accepted to Science.
   More details to come soon.

2. Change Point Localization and Inference in Dynamic Multilayer Networks
   F. Wang, K. Ritscher, Yik Lun Kei, X. Ma, O.H. Madrid Padilla
   International Conference on Learning Representations (ICLR) 2026
   PDF

3. Change Point Detection on A Separable Model for Dynamic Networks
   Yik Lun Kei*, H. Li*, Y. Chen, O.H. Madrid Padilla
   Transactions on Machine Learning Research (TMLR) 2025
   PDF, Video, library(CPDstergm)
   Funded by NSF DMS-2015489

4. Change Point Detection in Dynamic Graphs with Decoder-only Latent Space Model
   Yik Lun Kei, J. Li, H. Li, Y. Chen, O.H. Madrid Padilla
   Transactions on Machine Learning Research (TMLR) 2025
   PDF, Video

5. A Partially Separable Model for Dynamic Valued Networks
   Yik Lun Kei, Y. Chen, O.H. Madrid Padilla
   Computational Statistics & Data Analysis 2023
   PDF

6. YouRefIt: Embodied Reference Understanding with Language and Gesture
   Y. Chen, Q. Li, D. Kong, Yik Lun Kei, S. Zhu, T. Gao, Y. Zhu, S. Huang
   The IEEE International Conference on Computer Vision (ICCV) 2021 (Oral)
   PDF

Preprints

1. Decoder-only Clustering in Graphs with Dynamic Attributes
   Yik Lun Kei, O.H. Madrid Padilla, R. Killick, J. Wilson, X. Chen, R. Lund
   Under Review
   PDF

2. Confidence Interval Construction and Conditional Variance Estimation with Dense ReLU Networks
   C.M. Madrid Padilla*, O.H. Madrid Padilla*, Yik Lun Kei, Z. Zhang, Y. Chen
   Under Review
   PDF

3. Change Point Detection for Cell Populations Measured via Flow Cytometry
   Yik Lun Kei*, Q. Wang*, P. Parker, F. Ribalet, S. Hyun
   Under Review
   PDF
   Funded by NSF CAIG

In Progress

1. Clustering Auto-regressive Models on Weighted Graphs
   Yik Lun Kei, O.H. Madrid Padilla, R. Killick, J. Wilson, X. Chen, R. Lund
   2026+
   library(GraphClustAR)

Others

1. Online Multi-robot Deadlock Prediction with Conditional Variational Auto-Encoder for Sequences
   Yik Lun Kei, M. Zhang, C. Stolz, A. Aroor, Y. Zhang, Y. Gao
   Amazon Robotics Science Summit
   Work performed during internship at Amazon

* denotes equal contribution

Teaching

STAT 292: Advanced Topics in Statistics

STAT 266A: Data Visualization and Statistical Programming in R

STAT 132: Classical and Bayesian Inference

STAT 131: Introduction to Probability Theory

STAT 80A: Gambling and Gaming

STAT 280B: Seminars in Statistics

Industry Research Experience

Cisco
Software Engineer Intern (2024)

Amazon Robotics
Data Scientist Intern (2023)

UCLA Health
Senior Data Analyst (2017-2019)

Software

library(CPDstergm)
Github

• An R package to detect multiple change points in time series of graphs, using Separable Temporal Exponential-family Random Graph Model (STERGM). The optimization problem with Group Fused Lasso regularization on the model parameters is solved by Alternating Direction Method of Multipliers (ADMM).

library(GraphClustAR)
Github

• An R package for clustering node-level time series by jointly modeling temporal dynamics and network structure via Graph-fused LASSO regularization and autoregressive models.

Talks

Department of Statistics, University of California, Santa Cruz (Apr. 2024)
Department of Mathematics and Statistics, University of San Francisco (Feb. 2026)

Awards

UCLA Graduate Fellowship (2021-2023)
UCLA Summer Mentored Research Fellowship (2022)

Services

Journal/Conference Reviewer
NSF Grant Reviewer
Master Thesis Committee
Graudate Student Mentor