Math4ML Workshop

Machine Learning (ML) has motivated a lot of new research in mathematics. This research uses tools from both pure and applied mathematics. The workshop will showcase some of the recent mathematical research motivated by problems coming from ML and some of the mathematical developments leveraging tools from machine learning to improve scientific discovery and engineering applications.

These include but are not limited to

• PINNs (a neural networks method for numerically solving differential equations)

• Optimization for ML  (local vs global minima, convergence of SGD, IP and MIP formulations of deep neural networks, robust optimization, …)

• Recent developments in Statistical Learning ( generalization, over-parameterization, …) 

• Probabilistic approaches (Bayesian ML, transfer learning, …)

• Graph theory and related topics (Knowledge graphs, Graph Neural Networks, ...)

• Equation Learning (symbolic regression)

• Topological data analysis

The main goal of this workshop is to showcase the mathematical research around ML to the Deep Learning Indaba community but especially to the mathematicians in the community. This will hopefully inspire young African mathematicians to consider working on some of these problems. The long-term expected outcome is to grow the research community around the topics discussed in the workshop, which currently has few African members of the community.