Welcome to the third video of our lecture series on Data-Driven Models for Unsteady Fluid Flows. In this video, we delve into dimensionality reduction techniques, which are crucial for efficiently analyzing and modeling high-dimensional fluid flow data. From modal decomposition methods like Proper Orthogonal Decomposition (POD) and Dynamic Mode Decomposition (DMD) to neural networks and community-based reductions, we'll explore an array of techniques to make your data more manageable and insights more accessible.
🕒 Timestamps:
0:00 - Introduction
0:23 - Modal Decomposition Overview
3:11 - Proper Orthogonal Decomposition (POD)
15:04 - Dynamic Mode Decomposition (DMD)
31:36 - Spectral Proper Orthogonal Decomposition (SPOD)
35:38 - Extended Dynamic Mode Decomposition (EDMD)
39:11 - Neural Networks in Dimensionality Reduction
50:44 - Autoencoders for Fluid Flow Data
59:34 - Community-Based Reduction
1:04:10 - Cluster-Based Reduction
1:09:09 - Quick recap
This lecture series was recorded as part of the NSF AI Institute of Dynamic Systems. Check out dynamicsai.org/edu for more educational resources.
#FluidDynamics #DimensionalityReduction #POD #DMD #NeuralNetworks #DataDrivenModels