DIKU Bits: A visual tour of shape analysis

Speaker

Morten Akhøj Pedersen, PhD student in the Pioneer AI (P1AI) section

Title

A visual tour of shape analysis

Abstract

In the field of 'shape analysis' we construct a mathematical space in which each point represents a shape, e.g. a curve or a surface. This talk illustrates this space of shapes: what does the shortest path from a dog to a cat look like? How about a random walk starting from a butterfly? How can we do statistics and machine learning on a set of shapes? Answers to such questions are important in e.g. evolutionary biology, which is the focus of our Center for Computational Evolutionary Morphology.

Which courses do you teach?

I recently taught Mathematical Analysis and Probability Theory for Computer Scientists (MASD) and Introduction to data science. I have no further teaching plans this year. However, I (and others in our group) would be very happy to supervise projects!

Which technology/research/projects/startup are you excited to see the evolution of?

I'm excited about the development of open source large language models (e.g. Mistral 7B, LLama2) that can be run locally and privately. I'm also very interested in the further improvement of automatic differentiation tools, especially for more mathematically complicated operations (e.g. differentiating solutions to optimization problems or differential equations).