Summer Graduate School

DATES MAY CHANGE

The overarching goal of this summer school is to expose the students both to modern forms of unsupervised learning — in the form of geometrical and topological data analysis — and to supervised learning — in the form of (deep) neural networks applied to regression/classification problems. The organizers have opted for a lighter exposure to a broader range of topics. Using the metaphor of a meal, we are offering 2 + 2 samplers — geometry and topology for data analysis + theoretical and practical deep learning — rather than 1 + 1 main dishes. The main goal, thus, is to inspire the students to learn more about one or several of the topics covered in the school.

The expected learning outcomes for students attending the school are the following:

1. An introduction to how concepts and tools from geometry and topology can be leveraged to perform data analysis in situations where the data are not labeled.

2. An introduction to recent and ongoing theoretical and methodological/practical developments in the use of neural networks for data analysis (deep learning).

School Structure

Each of the two courses will have a daily lecture followed by a discussion / problem session.  Each afternoon will include a presentation or social event. 

Prerequisites

The lectures will be accessible to graduate students in mathematics or closely related fields who have taken courses (at least at the upper division level) in linear algebra, multivariate calculus, real analysis, differential geometry, and topology. Some programming experience with languages such as Python, R, Julia, Matlab, is also desirable.

For eligibility and how to apply, see the Summer Graduate Schools homepage

Due to the small number of students supported by MSRI, only one student per nominating institution will be eligible to be funded by MSRI.