Machine Learning and Dynamical Systems Seminar
The study of complex systems incorporates various approaches, including dynamical systems theory and machine learning. Dynamical systems theory, established in the 19th century by figures such as Poincaré and Lyapunov, focuses on understanding the qualitative behavior of systems through models. These models, typically expressed through Ordinary/Partial, Underdetermined (Control), Deterministic/Stochastic differential, or difference equations, approximate observed realities. They provide a mathematical framework for analyzing complex systems, although developing precise models for certain challenging domains like climate dynamics, brain function, biological systems, and financial markets is exceptionally challenging.
Conversely, machine learning centers on algorithms that improve performance with additional data, finding applications in areas such as computer vision, stock market analysis, speech recognition, recommender systems, and social media sentiment analysis. Machine learning excels in scenarios lacking explicit models but having ample measurement data, thus driving the growth of data-driven technologies across various fields. However, the theoretical aspects of machine learning are still underdeveloped.
The intersection of the theory of dynamical systems and machine learning offers a rich territory for exploration, and it naturally invites investigation into the synergy of these two domains in the following specific directions:
Machine Learning for Dynamical Systems: This involves analyzing dynamical systems using observed data rather than purely analytical methods, intending to expand the classical theory to more complex and challenging systems.
Dynamical Systems for Machine Learning: This approach examines machine learning algorithms as dynamical systems, applying dynamical systems theory tools to better understand these algorithms.
To facilitate collaboration and research in these areas, the Machine Learning and Dynamical Systems Seminar, initiated in October 2020, serves as an online platform. It offers research seminars, symposia, and reading groups at the intersection of these fields. The seminar, a part of the Special Interest Group on "Machine Learning and Dynamical Systems" (MLDSIG) at the Alan Turing Institute, which I co-lead with Prof. Robert Mackay, aims to bridge the gap between the theory of dynamical systems and machine learning.
We warmly welcome you to engage more closely with our initiatives and become an active part of our vibrant community. By exploring the resources outlined below, you will gain deeper insights into our collaborative efforts.
Our seminar is more than just a series of events; it's a growing community where cutting-edge ideas converge, fostering groundbreaking research at the intersection of these two dynamic fields.
Additional Resources:
For detailed information, [click here].
For a concise video overview, [click here].
Archival and Updates:
Past seminars are archived on our [Youtube Channel].
For regular updates about the Research Interest Group (RIG), please subscribe to our mailing list by completing the form available [here].
Symposia on Machine Learning and Dynamical Systems:
First Symposium on Machine Learning and Dynamical Systems, Imperial College London, Feb. 11-13, 2019
Upcoming Schedule: The schedule for future seminars is provided below.
