Name: Bolei Zhou
Title: Computer Vision for Indoor Scene Understanding
Description: Semantic visual understanding of the indoor environments goes beyond the single image recognition. With the wide access of scanning devices such as iPhone and Android mobile phones, it is convenient to scan the surrounding daily environments in 3D and recognize their semantic properties, such as the floor plan and the room types. However, when scaling up the visual scene understanding from single image to the real-world 3D surroundings, it remains difficult to scan, annotate, and understand their semantic properties.
Preferred qualifications: Applicants should have a good knowledge of Python programming, statistics, and linear algebra. It will be a huge plus if the applicants have experience in machine learning, deep learning, and have done some previous projects in PyTorch and deep learning

Name: Guy Van den Broeck
Title: Towards Probabilistic Programming
Description: This project will extend our current probabilistic programming language with novel language features, applications, or probabilistic inference algorithms.
Preferred qualifications: Basic Probability and Programming

Name: Sanjukta Krishnagopal + Mason A. Porter
Title: 'Information Equity' in Social Networks
Description: Social networks, in the form of graphs and their generalizations, encode interactions between entities in population. Each entity (e.g., an individual person) is a node of a network, and connections between these nodes are edges of the network. The structure of a social network plays an important role in the flow of information on it. In this project, the student will study how different network structures (e.g., networks with communities, networks with core-periphery structure, etc.) influence the propagation of information (and hence opportunities for the nodes) in networks. The student will investigate optimal intervention strategies to 'seed' information at carefully chosen individuals to maximize information access (for the purpose of 'information' equity). To do this, the student will simulate mathematical models of information spread and will compare how intervention strategies to maximize information access of minority groups differ for different network structures (and, time permitting, for different choices of the information-spreading model).
Preferred qualifications: Programming (preferably in Python) and upper-division linear algebra. Experience with probability and networks is a plus.