Sparse Inference, and Network and Text Analysis
STAT 236
Subject & Catalog Number
Course Information
Description
High dimensional data analysis is a recent interdisciplinary research area of Statistics, Genetics and Genomics, Engineering, and several other scientific areas. It addresses an array of challenging problems that of contemporary interest, and research in this area has been very active in the past decade.
This course aims to provide a systematic introduction to various topics in high dimensional data analysis, focusing on large-scale sparse learning, and network and text data analysis. Large-scale sparse learning: Sparsity is a universal phenomenon in modern high dimensional data. Sparse structures are observed in many application settings and have many different forms, such as parameter sparsity, graph sparsity, eigenvalue sparsity, and so on. Exploring sparsity has become a common strategy in data analysis and has largely reshaped classical multivariate statistics problems. This course will investigate classical problems such as multiple testing, linear regression, classification and clustering, under the modern sparse settings. For each problem, the course discusses recent statistical methods for taking advantage of sparsity, and introduces the
theoretical framework for analyzing these methods.
Network and text data analysis: Social networks and text documents are unconventional data types. This course introduces statistical models and methods for analyzing such type of data. Topics for network data analysis include community detection, mixed membership estimation, link prediction, and dynamic network modeling. Topics for text data analysis include topic modeling, word embedding, information retrieval, and sentiment analysis.
Recommended Prep
This course is mainly designed for graduate students in Statistics. The prerequisites are Statistical Inference 1&2 (students can take the prerequisites simultaneously with this course). Graduate students from other departments (CS, Biostatistics, Economics, etc.) can also take this course, if they have taken statistics-related courses which have a significant amount of content in mathematical statistics. Undergraduate students please consult the instructor before enrolling.
Available for Harvard Cross Registration
