M'AI: Neural ODEs 1000-1M25MNO

This course introduces the mathematical foundations of Neural Ordinary Differential Equations (Neural ODEs) - a groundbreaking approach combining ordinary differential equations theory with deep learning. Unlike traditional discrete neural networks, Neural ODEs model the learning process as continuous dynamics described by differential equations.Topics Covered:

1. Theoretical Foundations:

* Differential equations in machine learning
* Problem formulation as ODEs (adjoint method)
* Comparison with classical NN architectures

2. Numerical Aspects:

* Efficient equation integration (adaptive solvers)
* Backpropagation through ODE solvers

3. Applications:

* Modeling dynamic processes
* Generative flow models (Continuous Normalizing Flows)

Course Format: The course blends rigorous mathematical treatment with practical implementation aspects, demonstrating how modern mathematics becomes the language of artificial intelligence. Lectures will provide theoretical foundations;Tutorials will focus on analytical approaches to Neural ODEs; Multi-day offsite workshops (outside Warsaw) are planned pending funding availability.