Cindy Orozco Bohorquez
|
Ph.D. Candidate in Computational and Mathematical Engineering
Stanford University |
|
Cindy is a PhD candidate at the Institute for Computational and Mathematical Engineering (ICME) at Stanford. Her work combines modern tools of data analysis and optimization with traditional numerical analysis results.
Cindy has taught data science to a broad audience, including high school students from all parts of Colombia in a program called Clubes de Ciencia, as well as Silicon Valley engineers and managers in the ICME Summer workshops. She is originally from Colombia, where she did a bachelor's in civil engineering and mathematics at Universidad de los Andes, and she also hold a master's in applied mathematics from King Abdullah University of Science and Technology, in Saudi Arabia. Cindy is also one of the cofounders of WiMSCE: Women in Mathematics, Statistics and Computational Engineering at Stanford, a graduate student group that brings together people from different departments with the goal of creating a community and learning about challenges and perspectives from women in these fields.
|
Keynote: What does it mean to have a robust algorithm? Looking through the lenses of point-set registration.
When we are fitting a model to a data set, we would like to have the certainty that we are recovering the ground truth. In most cases, we do not obtain a good fit at first, and we need to take a decision of what to do next. Some increase the sample size. Some change the optimization algorithm or the mathematical model and assumptions. Some conclude that there is not enough signal and resample with a higher precision or from a different population. Which of these choices is the right one?
In this talk, we study this question for a classical problem in computer graphics and satellite communication called point-set registration. We focus on the special case of recovering the rotation that aligns two data sets that belong to the d-dimensional sphere. Traditionally, under the assumption of low noise, this problem is effectively solved using linear algebra. When the low-noise assumption is no longer valid, other algorithms have been proposed. We combine results from statistics, optimization, and differential geometry, to compare the solutions given by these algorithms. As a result, we provide a direct mapping between the rate of success of each algorithm and the peculiarities of different datasets.
In this talk, we study this question for a classical problem in computer graphics and satellite communication called point-set registration. We focus on the special case of recovering the rotation that aligns two data sets that belong to the d-dimensional sphere. Traditionally, under the assumption of low noise, this problem is effectively solved using linear algebra. When the low-noise assumption is no longer valid, other algorithms have been proposed. We combine results from statistics, optimization, and differential geometry, to compare the solutions given by these algorithms. As a result, we provide a direct mapping between the rate of success of each algorithm and the peculiarities of different datasets.