Can AI solve equations?
17.10.2023Solving equations is hard. So hard that scientists have been developing new methods to do so for the past 2000 years and are still doing so. Why not use neural networks to do it?
Solving equations is hard. So hard that you almost failed your maths classes because of it. So hard that scientists have been developing new methods to do so for the past 2000 years and are still doing so. Why not use neural networks to do it? This bold question led to this exciting thesis project.
Harnessing the power of modern machine learning, this project aims at combining Physics-Informed Neural Networks (PINNs) with the concept of hypernetworks. PINNs, known for their ability to incorporate physical principles into deep learning models, offer a promising approach for solving complex differential equations. Hypernetworks, on the other hand, provide a unique framework for generating network weights dynamically, enabling the neural network to adapt to different starting conditions.
The potential implications of this research are profound. Imagine a world where complex physical systems can be modeled and solved with unprecedented accuracy and speed. Applications span from climate modeling and energy optimization to drug discovery and aerospace engineering. This thesis project is only the first step on this way, but even the longest journey begins with a first step!
Supervisor: Andrzej Dulny
