j nathan kutz
play

J. Nathan Kutz Department of Applied Mathematics University of - PowerPoint PPT Presentation

DATA-DRIVEN MODEL DISCOVERY AND COORDINATE EMBEDDINGS FOR PHYSICAL SYSTEMS ICERM 2019 J. Nathan Kutz Department of Applied Mathematics University of Washington Email: kutz@uw.edu Web: faculty.washington.edu/kutz Model Discovery Steven


  1. DATA-DRIVEN MODEL DISCOVERY AND COORDINATE EMBEDDINGS FOR PHYSICAL SYSTEMS ICERM 2019 J. Nathan Kutz Department of Applied Mathematics University of Washington Email: kutz@uw.edu Web: faculty.washington.edu/kutz

  2. Model Discovery Steven Brunton Mechanical Engineering University of Washington Joshua Proctor Institute for Disease Modeling

  3. Mathematical Framework State-space Dynamics Parameters x Stochastic effects Dynamics Measurement Measurement noise Measurement model

  4. What Could the Right Side Be? Limited by your imagination 2 nd degree polynomials

  5. Sparse Identification of Nonlinear Dynamics (SINDy)

  6. Identifying ROMs

  7. Discovery Innovations Schaeffer et al -- corrupt data, PDEs, integral formulation, convergence Dongbin Xiu & co-workers (2018) – Sampling strategies Guang Lin & co-workers (2018) -- Uncertainty Metrics Hod Lipson and co-workers (2006) — Symbolic/genetic regression Karniadakis, Raissai, Perdikaris …. — Neural Nets Zheng, Askham, Brunton, Kutz & Aravkin (2018) – SR3 sparse relaxed regularized regression (for SINDy, LASSO, CS, TV, Matrix Completion …)

  8. Manifolds and Embeddings Observables & Coordinates

  9. Bernard Koopman 1931 Mezic (2004) Coifman, Kevrekidis, co-workers

  10. Koopman Invariant Subspaces Brunton, Proctor & Kutz, PLOS ONE (2018)

  11. Burgers’ Equation Cole-Hopf Kutz, Proctor & Brunton, Complexity (2018)

  12. Dynamic Mode Decomposition Travis Askham Askham & Kutz, SIADS (2018)

  13. Approximate Dynamical Systems Linear dynamics (equation-free) Eigenfunction expansion Least-square fit

  14. DMD with Control Input Input Snapshots DMD generalization Proctor, Brunton & Kutz, SIADS (2016)

  15. Koopman vs DMD: All about Observables!

  16. Neural Nets

  17. The Zoo NN Zoo

  18. NNs for Koopman Embedding Bethany Lusch Lusch et al. Nat. Comm (2018)

  19. Handling the Continuous Spectra NNs for Koopman Embedding Bethany Lusch Lusch, Kutz & Brunton, arxiv (2017) Lusch et al. Nat. Comm (2018)

  20. The Pendulum

  21. Flow Around a Cylinder

  22. Autoencoder + SINDy Kathleen Champion

  23. Fast Learning Charles Delahunt

  24. Moth Olfactory System

  25. Learning New Odors

  26. Sparsity for Learning

  27. Rapid Learning in NNs

  28. Comparisons

  29. Decoder Networks

  30. Structure of Mapping Linear Maps: SVD (left singular vector) defines layer Erichson, Mathelin, Brunton, Mahoney & Kutz (2019)

  31. Mathematical Framework State space Measurements Mapping Approximate the full state space from limited measurements Optimization

  32. Linear Maps Singular value decomposition Data Linear measurements H Optimize (least-squares)

  33. Optimal Placement Point measurements Optimal Sensors via QR pivots Manohar, Kutz & Brunton (2018) IEEE Control Systems Magazine Clark, Askham, Brunton & Kutz (2019) IEEE Sensors

  34. Nonlinear Mapping General Form: Compositional Layers Universal Approximators: Hornik 1990

  35. Shallow Layer Mapping Two Layers Composition Erichson, Mathelin, Brunton, Mahoney & Kutz (2019) SIAM

  36. Activation Functions

  37. Linear vs Nonlinear Maps Improved Interpretability of Modes

  38. Improved Performance

  39. Robustness to Noise

  40. Improved Performance

  41. Super Resolution Analysis

  42. Conclusion Model Discovery: Sparse regression provides parsimonious dynamical models Coordinates: Learning Koopman embeddings can provide optimal basis for dynamics Neural Networks: Structure and function matter - Connect discovery and coordinates - Structure can lead to fast (one-shot) learning with limited data - Discovery of improved coordinate embeddings through decoding COMING SOON: A multi scale physics challenge set

Recommend


More recommend