Summary of Sparse Identification of Nonlinear Dynamics (SINDy): Sparse Machine Learning Models 5 Years Later!

00:00:00 - 00:20:00

Sparse identification of nonlinear dynamics (SINDy) is a powerful tool used to learn the true dynamics of a system from observational data. This technique is applicable to a wide range of systems, including plasma, fluid flow, and economic models. The cindy algorithm augments the right-hand side of possible model terms to include nonlinear terms, in order to find the fewest terms in a library of candidate dynamics that describe the x dot, y dot, and z dot equations. This allows for models that are more like the Lorentz 1963 model, which is able to describe the rich dynamics of the atmosphere.

• 00:00:00 SINDy is a sparse identification algorithm for learning dynamical systems models from time series data. After introducing the algorithm and its philosophy, the video goes on to discuss how the algorithm works and what has been learned in the past five years.
• 00:05:00 This 1-part video series will teach viewers how to compute sparse identification models, and what pitfalls to avoid when doing so. The series will also focus on some of the applications of sparse identification models, such as modeling fluid flows, the human brain, and ecological systems.
• 00:10:00 The cindy algorithm augments the right-hand side of possible model terms to include nonlinear terms, in order to find the fewest terms in a library of candidate dynamics that describe the x dot, y dot, and z dot equations. This allows for models that are more like the Lorentz 1963 model, which is able to describe the rich dynamics of the atmosphere.
• 00:15:00 Sparse identification of nonlinear dynamics (SINDy) is a powerful tool used to learn the true dynamics of a system from observational data. This technique is applicable to a wide range of systems, including plasma, fluid flow, and economic models.
• 00:20:00 This video provides a schematic overview of sparse identification of nonlinear dynamics (SINDy). Challenges associated with this method include finding the sparsest pde that describes the data, designing a library of candidate dynamics terms, and optimizing the model for sparse solutions. Success stories from fields such as fluid dynamics are also discussed.