Research

Realizing the need for a solid and diverse educational foundation for deeper statistical studies, I self-studied advanced topics in Statistics and Machine Learning from my freshman year. In my sophomore spring, I attended seminars on emerging subjects and independently studied Optimization Theory, honing my intuition and expertise. I also believe a solid foundation in Computer Science, could enhance my programming skills and breadth of knowledge, integrating computational thinking into my future research. Therefore, I took an additional 20 credits of core courses at the School of Computer Science. During my undergraduate studies, my research experience primarily focused on machine learning methods, statistical network analysis, and data science. These scientific research experiences have produced some results, including but not limited to some academic papers and software.

During my undergraduate years, I also explored various research areas, hoping to delve into more fields in the future to address genuinely interesting and meaningful problems. I am particularly interested in the development and application of statistical and data science methodologies, as well as interdisciplinary research between statistics and fields such as physics and biology. I am also very interested in network analysis, machine learning theory and algorithms, and optimization. Progress in these fields requires a fusion of a solid mathematical foundation, a statistical lens, and computational thinking. In-depth exploration of different disciplines can broaden one’s perspective, equipping one with core technologies and offering more choices for the future.

Research Interests

• Interdisciplinary Study
• Network Analysis
• Machine Learning
• Optimization
• Data Science
• Deep Learning Theory

Research Experience

Time Series Forecasting Framework: Capturing Underlying Volatility Information (USTC, China)

Team Leader, University of Science and Technology of China (Dec. 2022 - Jul. 2023)

• Investigated contemporary time series hybrid forecasting models, proposing a novel time series decomposition-ensemble forecasting framework that integrates traditional statistical model GARCH with neural networks.
• Implemented diverse time series forecasting algorithms, evaluated their predictive efficacy on the German Consumer Price Index dataset, and explained the superiority of our proposed framework.
• Compared the experimental results using MAE, RMSE and MAPE metrics, revealing the superior performance of our model through substantial reductions in MAE, RMSE and MAPE values.

SCORE for Community Detection in Multi-layer Networks with Covariates (USTC, China)

Team Leader, University of Science and Technology of China (Mar. 2023 - Present)

• Introduced a novel multi-layer degree-corrected stochastic block model and developed the MSCORE algorithm to model the Multi-layer network clustering problem with covariates.
• Demonstrated the superiority of the MSCORE algorithm by the performance of ARI in different simulation experiments, explained the core principles of the algorithm.
• Proposed a novel multivariate normal-ratio distribution and derived its density function from different perspectives, investigated the basic theory of this distribution.

Network Reconstruction with Dependent Connectivity from Rich but Noisy Data (USTC, China)

Core Participant, University of Science and Technology of China (Mar. 2023 - Oct. 2023)

• Participated in proposing a generalized EM algorithm to restore the network structure and led the project’s simulation experiments, improving computational efficiency through data sparsification and integration of the Rcpp package.
• Investigated fMRI images and derived correlation matrices from blood oxygen level time series, reconstructed matrices and grouped regions of interest in the human brain to obtain structural information, aiding in Alzheimer’s diagnosis.
• Provided strong consistent estimates of the reconstructed network and communities under the setting where both the number of nodes and each pairwise measurements tend to infinity.

Network Clustering: Several Feasible Extensions to the Network Embedding Model (USTC, China)

Team Leader, University of Science and Technology of China (Jul. 2023 - Present)

• Conducted research on the Network Embedding Model and organized a research group for related discussions.

• Extended Network Embedding Model so that it can be applied to highly degree-heterogeneous networks, studied the impact of different penalty functions on the effect of the extended model.

• Completed the simulation experiment and evaluated our method against various clustering algorithms, including SCORE, Classic Network Embedding Model, Spectral Clustering, etc.

CODIA Intelligent Diagnosis Development Team of BDAA Laboratory (USTC, China)

Core Participant, University of Science and Technology of China (Jul. 2022 - Jan. 2023)

• Designed and developed the front-end web pages using the Vue framework, offering users exercise recommendations and personal ability assessments.
• Investigated literature related to intelligent diagnosis and recommendation systems, reproduced associated algorithms and implemented API interfaces for these algorithms.
• Developed and launched the CODIA website with our team. (This project won the championship in the ”Spark Cup” Cognitive Large Model Scene Innovation Competition. Congratulations!) CODIA website link

Network Analysis and Time Series Theory Reading Group (USTC, China)

Advisor: Prof. Yu Chen, Department of Statistics and Finance, USTC (Nov. 2022 - Present)

• Conducted in-depth network analysis and time series literature review exploration, and actively participated in workshops to exchange research ideas with peers and supervisors.
• Delivered eight presentations as the principal presenter, introducing combined models related to Time Series, as well as algorithms for Network Clustering, Tucker Decomposition, and others.
• Guided discussions and actively proposed research ideas in each group meeting, evaluating the feasibility of research ideas through theoretical analysis and simulation experiments.

Neural Networks and Cognitive Diagnosis Seminar (USTC, China)

Advisor: Prof. Qi Liu, Department of Data Science, USTC (Jul. 2022 - Feb. 2023)

• Actively participated in and discussed the research on cognitive diagnostic models, offering insights and reporting on research progress.
• Systematically read and studied several books (e.g. Hands-on Deep Learning, Statistical Learning Methods), as well as some review papers and popular neural network models.
• Delivered presentations on articles related to neural cognitive diagnostic models and led the research group to understand the core principles of the model and discuss research ideas.