Zexi Huang (黄泽熙)
Machine Learning Scientist at TikTok Recommendation

Welcome, my name is Zexi Huang. I am a Machine Learning Scientist at TikTok Recommendation. I obtained my PhD in Computer Science at the Department of Computer Science, University of California, Santa Barbara (UCSB) in April 2023. Prior to that, I received my Bachelor's in Computer Science and Technology with the highest honor at Yingcai Honors College, University of Electronic Science and Technology of China (UESTC), in 2018.

 

My research focuses on machine learning and data mining on information-rich data. At TikTok, I work on developing state-of-the-art machine learning solutions for the billion-scale recommendation system of TikTok Live. My PhD dissertation at Dynamic Networks: Analysis and Modeling Lab, UCSB, is on representation learning for information-rich graphs, with Prof. Ambuj Singh as my advisor. Previously, I had multiple applied science internships at Books Tech, Amazon, where I leveraged graph-based machine learning techniques to solve large-scale industry problems including fraud detection, inventory management, and content discovery. I also interned at Computational Intelligence Lab, Nanyang Technological University, working with Prof. Sinno Jialin Pan on transfer learning framework for community detection in multiplex networks, as an undergraduate.

News

Feb. 2024

We presented our work on GNN-based attributed graph clustering at AAAI'24!

May. 2023

I joined TikTok as a Machine Learning Scientist and will be working on TikTok Live Recommendation!

Apr. 2023

I defended my PhD dissertation on Learning Representations for Information-rich Graphs!
Education

Sep. 2018 - Apr. 2023

College of Engineering, University of California, Santa Barbara
Doctor of Philosophy and Master of Science in Computer Science
  • Course GPA: 4.0/4.0. UCSB Computer Science Outstanding Scholar Fellow (top 4 out 63 PhD students).
  • Advisor: Prof. Ambuj Singh. Dissertation: Learning Representations for Information-rich Graphs.

Sep. 2014 - Jun. 2018

Yingcai Honors College, University of Electronic Science and Technology of China
Bachelor of Engineering in Computer Science and Technology (Honors)
  • GPA: 3.96/4.0, Avg. Score: 92.79/100 (1/90 in freshman year, 1/87 in sophomore year, 1/93 in junior year).
  • Advisors: Prof. Junming Shao and Prof. Sinno Jialin Pan. Thesis: Transfer Learning for Community Detection in Multiplex Networks.

Feb. 2016 - Jun. 2016

College of Electrical and Computer Engineering, National Chiao Tung University
Exchange Student
  • GPA:4.0/4.0, Avg. Score: 96.44/100, Straight A+.
Work Experience

May. 2023 - Present

Machine Learning Scientist, Live Recommendation, TikTok
Billion-scale Machine Learning for TikTok Live Recommendation
  • Owned the iterations of the core ranking models of the recommendation system for TikTok Live, leveraging representation learning, multi-task learning, knowledge distillation, and sequence modeling, with +2% user watch-live duration gains from online A/B experiments.
  • Developed the host go-live model that captures the relationship between watch-live and go-live with causal inference and uplift modeling to motivate authorized hosts to go-live, achieving +2% user go-live penetration in online A/B experiments.

Jun. 2022 - Sep. 2022

Applied Scientist Intern, Books Tech, Amazon
Stochastic Inventory Management for Print-On-Demand and Graph-based Text Classification for Content Intelligence
  • Developed a stochastic inventory management model based on dynamic programming for Amazon’s Print-On-Demand business, optimizing the ordering strategy for 42.4M units of books and realizing an annual saving of $10.4M.
  • Designed and implemented a graph-based NLP model for long text embedding and classification using graph neural networks, leading to superior performance to state-of-the-art transformer-based models for Kindle book contents.

Jun. 2021 - Sep. 2021

Applied Scientist Intern, Books Tech, Amazon
Graph-based Fraud Detection in Kindle Direct Publishing
  • Designed graph-based machine learning models for fraud detection based on multi-modal signals in Kindle Direct Publishing.
  • Implemented the models in an end-to-end fashion and deployed them to graphs with millions of nodes and billions of edges in production.
  • Validation results show that the models can surface fraud rings undetected by existing processes with an estimated annual value of $2.4M.

Jun. 2020 - Sep. 2020

Applied Scientist Intern, Books Tech, Amazon
Graph-based Fraud Detection in Kindle Direct Publishing
  • Proposed to augment existing fraud detection methods with graph-based machine learning models for Kindle Direct Publishing.
  • Designed and implemented various heuristics and an embedding framework for attributed heterogeneous multiplex networks.
  • The models are deployed into production and results show that they surface up to 15 times more fraud compared to the existing processes.

Sep. 2017 - Feb. 2018

Research Intern, Computational Intelligence Lab, Nanyang Technological University
Transfer Learning for Community Detection in Multiplex Networks
  • Proposed to refine community detection results in some layers with transferred knowledge from other layers in multiplex networks.
  • Designed a representation-based community detection framework and implemented it with an extended symmetric NMF approach.
  • Our algorithm outperforms other representation-based community detection algorithms, especially when the target layer is noisy.
Publications

[AAAI-2024] Aritra Bhowmick, Mert Kosan, Zexi Huang, Sourav Medya, Ambuj Singh, Sourav Medya. DGCLUSTER: A Neural Framework for Attributed Graph Clustering via Modularity Maximization. AAAI Conference on Artificial Intelligence, 2024. [Code]

[WSDM-2023] Zexi Huang*, Mert Kosan*, Sourav Medya, Sayan Ranu, Ambuj Singh. Global Counterfactual Explainer for Graph Neural Networks. ACM International Conference on Web Search and Data Mining, 2023. (*: equal contribution) [Code] [Slides] [Talk]

[WSDM-2022] Zexi Huang, Arlei Silva, Ambuj Singh. POLE: Polarized Embedding for Signed Networks. ACM International Conference on Web Search and Data Mining, 2022. [Code] [Poster] [Slides] [Talk]

[KDD-2021] Zexi Huang, Arlei Silva, Ambuj Singh. A Broader Picture of Random-walk Based Graph Embedding. ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021. [Code] [Poster] [Slides] [Talk] [Talk (in Chinese)]

[Preprint] Zexi Huang, Mert Kosan, Arlei Silva, Ambuj Singh. Link Prediction without Graph Neural Networks. arXiv preprint arXiv:2305.13656, 2023. [Code]

[Preprint] Wei Ye, Zexi Huang, Yunqi Hong, Ambuj Singh. Graph Neural Diffusion Networks for Semi-supervised Learning. arXiv preprint arXiv:2201.09698, 2022. [Code]

Research Experience

Apr. 2021 - Apr. 2023

Multiscale Community Detection Based on Pointwise Mutual Information
Advisor: Prof. Ambuj Singh, Collaborator: Manu Kondapaneni, Arlei Silva
  • Developed a stability optimization algorithm for multiscale community detection based on the pointwise mutual information similarity.
  • Preliminary experiments show that the proposed algorithm can uncover the natural scales for different communities in the graph.

May. 2021 - Feb. 2023

Link Prediction without Graph Neural Networks
Advisor: Prof. Ambuj Singh, Collaborators: Mert Kosan, Arlei Silva
  • Scrutinized the training and evaluation of link prediction methods and identify their limitations in handling class imbalance.
  • Proposed a novel topology-centric framework that combines graph learning, topological heuristics, and an N-pair loss for link prediction.
  • Results showed that the proposed method is 145% more accurate and trains/infers 11/6,000 times faster than the state-of-the-art methods.

Oct. 2021 - Aug. 2022

Global Counterfactual Explanation for Graph Neural Networks
Advisor: Prof. Ambuj Singh, Collaborators: Mert Kosan, Sourav Medya, Sayan Ranu
  • Formulated the novel problem of global counterfactual reasoning/explanation of graph neural networks for graph classification.
  • Proposed GCFExplainer, the first global explainer powered by vertex-reinforced random walks on an edit map with a greedy summary.
  • Results showed that GCFExplainer not only provides crucial high-level insights but also outperforms existing methods in recourse quality.

Oct. 2020 - Aug. 2021

Signed Embedding for Polarized Graphs
Advisor: Prof. Ambuj Singh, Collaborator: Arlei Silva
  • Designed a novel polarization measure for signed graphs and showed that existing methods fail in polarized signed link prediction.
  • Proposed a polarized embedding algorithm that captures both topological and signed similarity jointly via signed autocovariance.
  • Extensive experiments showed that the proposed model outperforms state-of-the-art methods by up to one order of magnitude.

Apr. 2020 - Oct. 2021

Multiscale Graph Convolution via Neural Diffusions
Advisor: Prof. Ambuj Singh, Collaborator: Wei Ye, Yunqi Hong
  • Interpreted the layer-wise propagation rule of GCN from the perspective of power iteration and analyzed its converging process.
  • Designed a novel GCN architecture that learns to aggregate multiscale information based on graph diffusions with a neural network.
  • Illustrated the effectiveness and efficiency of the proposed model by extensive comparative studies with state-of-the-art methods.

Sep. 2018 - Feb. 2021

Graph Representation Learning Based on Random-walks
Advisor: Prof. Ambuj Singh, Collaborator: Arlei Silva
  • Presented a unified view of embedding, covering different random-walk processes, similarity metrics, and embedding algorithms.
  • Showed both theoretical and empirical evidence of the superiority of the novel autocovariance embedding in link prediction.
  • Illustrated ways to exploit the multiscale nature of random-walk similarity to further optimize embedding performance.

Jan. 2020 - Aug. 2020

Prospect Theory for Group Decision Making Dynamics
Advisor: Prof. Ambuj Singh, Collaborator: Mert Kosan
  • Extended the Prospect Theory to model group-level risky decision making dynamics with an interpersonal influence system.
  • Results on two human-subject experiments show that the group behavior shifts towards consensus and is explained by the influence.

Jul. 2016 - Aug. 2017

Overlapping Community Detection Based on Game Theory-incorporated Label Propagation Dynamics
Advisor: Prof. Junming Shao
  • Proposed an intuitive algorithm for fast overlapping community detection in networks.
  • Introduced gain and loss functions from game theory to model the intrinsic dynamics between nodes.
  • Results are comparable to state-of-the-art algorithms.
Teaching Experience

2020 - 2021

Lead Teaching Assistant
Department of Computer Science, UCSB

Fall 2019

Teaching Assistant, CS 130A: Data Structures and Algorithms
Department of Computer Science, UCSB

Spring 2019

Co-designer and Instructor, Machine Learning Workshops
LMU/UCSB Junior Nanotech Network PhD Student Exchange and Symposium

Winter 2019

Teaching Assistant, CS 8: Introduction to Computer Science
Department of Computer Science, UCSB

Fall 2018

Teaching Assistant, CS 174A/174N: Fundamentals of Database Systems
Department of Computer Science, UCSB
Honors & Awards

Feb. 2023

Top 10 best papers of WSDM'23 and best paper in the MLoG workshop
Top 10 out of 123 accepted papers and 690 valid submissions

Feb. 2022

WSDM NSF Travel Award
$300

Sep. 2020

UCSB Computer Science Lead Teaching Assistant Fellowship
$3,225

Sep. 2018

Computer Science Outstanding Scholar Fellowship at UCSB
6%, top four out of 63 admitted PhD students, $10,000

Sep. 2018

Computer Science Academic Excellence Fellowship at UCSB
$2,000

Jun. 2018

Outstanding Bachelor Thesis Award of UESTC
10%

Jun. 2018

Tang Lixin Scholarship for Studying Abroad
$1,453

Dec. 2017

The Most Outstanding Students Award of UESTC
0.2%, top 10 out of 5,000 seniors, $2,180

Dec. 2017

Honorary Graduate of Sichuan Province
1%, top one in Yingcai Honors College

Dec. 2017

Honorary Graduate of UESTC
10%, top twelve in Yingcai Honors College

Oct. 2017

National Scholarship of China
0.2%, top two in Yingcai Honors College, $1,162

May. 2017

National English Competition for College Students of China 2017
0.1%, Special Prize in National Final (Level C, for Non-English professionals)

Dec. 2016

People's Scholarship of UESTC
4%, Special Class, top four in Yingcai Honors College, $436

Dec. 2015

Tang Lixin Scholarship of Yingcai Honors College
1%, First Prize, top one in Yingcai Honors College, $4,359

Dec. 2015

Tang Lixin Scholarship of UESTC
0.2%, $1,453 per year until graduation

Dec. 2015

The 7th Chinese Mathematics Competitions
0.05%, 1st out of 1819 selected candidates in Sichuan Province
Academic Services
Registration Chair
  • KDD’23
Program Committee
  • AAAI’23-24
  • KDD’22
  • SDM’22
Reviewer
  • TNNLS’23
  • Neural Networks’23
  • TIST’22-24
  • TKDD’21-24
  • KDD’20-21
  • WebConf’21
Representative
  • Graduate Affairs Committee, Department of Computer Science, UCSB
General Judge
  • SB Hacks V, VII, VIII Hackathon

Last updated on May 5, 2024