Processing math: 28%
+ - 0:00:00
Notes for current slide
Notes for next slide

Persistent Homology of Coarse-Grained State Space Networks

Max Chumley

Mechanical Engineering

Michigan State University

10/14/23

1 / 24

Acknowledgements

——————————————————————————————

This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550- 30 22-1-0007

——————————————————————————————

2 / 24

Pipeline

3 / 24

Pipeline

3 / 24

Pipeline

3 / 24

Pipeline

3 / 24

Pipeline

3 / 24

Motivation - Ordinal Partition Networks (OPN)

4 / 24

Motivation - OPN Lorenz System

  • Lorenz System:

    dxdt=σ(yx),dydt=x(ρz),dzdt=xyβz

  • Timeseries: x(t)
5 / 24

Motivation - OPN Lorenz System

  • Lorenz System:

    dxdt=σ(yx),dydt=x(ρz),dzdt=xyβz

  • Timeseries: x(t)

  • Periodic: σ=10.0, β=83, ρ=100

5 / 24

Motivation - OPN Lorenz System

  • Lorenz System:

    dxdt=σ(yx),dydt=x(ρz),dzdt=xyβz

  • Timeseries: x(t)

  • Periodic: σ=10.0, β=83, ρ=100

  • Chaotic: σ=10.0, β=83, ρ=105
5 / 24

Motivation - OPN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1, b=0.2, c=14

  • Timeseries: z(t)

6 / 24

Motivation - OPN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1, b=0.2, c=14

  • Timeseries: z(t)

6 / 24

Motivation - OPN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1, b=0.2, c=14

  • Timeseries: z(t)

6 / 24

Motivation - OPN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1, b=0.2, c=14

  • Timeseries: z(t)

6 / 24

Motivation - OPN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1, b=0.2, c=14

  • Timeseries: z(t)

6 / 24

Motivation - Undesired Transitions

7 / 24

Coarse-grained State Space Networks (CGSSN)

8 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

Network Formulation

9 / 24

CGSSN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1

  • b=0.2
  • c=14

  • Timeseries: z(t)

10 / 24

CGSSN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1

  • b=0.2
  • c=14

  • Timeseries: z(t)

10 / 24

CGSSN with Noise

  • Periodic Rossler System

    dxdt=yz,dydt=x+ay,dzdt=b+z(xc)

  • a=0.1

  • b=0.2
  • c=14

  • Timeseries: z(t)

10 / 24

Distance Matrix

11 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=ePw(e)
12 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=ePw(e)

D(a,b)=min

12 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=\sum_{e\in P}w(e)

D(a,b)=\min_{P}C(P)

  • Unweighted: w(e)=1~\forall~e\in P
12 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=\sum_{e\in P}w(e)

D(a,b)=\min_{P}C(P)

  • Unweighted: w(e)=1~\forall~e\in P

12 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=\sum_{e\in P}w(e)

D(a,b)=\min_{P}C(P)

  • Unweighted: w(e)=1~\forall~e\in P

12 / 24

Graph Dissimilarity Measures - Shortest Path

  • Shortest Path: C(P)=\sum_{e\in P}w(e)

D(a,b)=\min_{P}C(P)

  • Unweighted: w(e)=1~\forall~e\in P

12 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}
13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

  • Shortest Weighted Path: D(a,b)=|\min_P C'(P)|
13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

  • Shortest Weighted Path: D(a,b)=|\min_P C'(P)|
13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

  • Shortest Weighted Path: D(a,b)=|\min_P C'(P)|

13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

  • Shortest Weighted Path: D(a,b)=|\min_P C'(P)|

13 / 24

Graph Dissimilarity Measures - Weighted Shortest Path

  • Weighted Shortest Path: C'(P)=\sum_{e\in P}\frac{1}{w(e)}

D(a,b)=\sum_{e\in P}w(e)

  • Shortest Weighted Path: D(a,b)=|\min_P C'(P)|

13 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Graph Dissimilarity Measures - Diffusion Distance

  • Diffusion Distance: P(i,j)=\frac{A(i,j)}{\sum_{k=1}^{|V|}A(i,k)}

    • A: Adjacency matrix with no self loops
    • P: Probability of transitioning from i to j
  • t-random walk: P^t(a,b)

  • Lazy Transition Probability: \widetilde{\mathbf{P}} = \frac{1}{2}\left[\mathbf{P}(a, b) + \mathbf{I}\right]

d_t(a,b) = \sqrt{ \sum_{c \in V} \frac{1}{\mathbf{d}(c)} { \left[ \widetilde{\mathbf{P}}^t (a,c) - \widetilde{\mathbf{P}}^t (b,c) \right] }^2 }

14 / 24

Network Distances Summary

15 / 24

Persistent Homology

16 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Persistence of Networks

17 / 24

Sine Wave Example

18 / 24

State Identification

19 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Empirical Testing of 23 Dynamical Systems

20 / 24

Dynamic State Identification Results

21 / 24

Experimental Results - Periodic

22 / 24

Experimental Results - Periodic

22 / 24

Experimental Results - Periodic

Myers, Audun, and Firas A. Khasawneh. "Dynamic State Analysis of a Driven Magnetic Pendulum Using Ordinal Partition Networks and Topological Data Analysis."

22 / 24

Experimental Results - Chaotic

23 / 24

Experimental Results - Chaotic

23 / 24

Experimental Results - Chaotic

Myers, Audun, and Firas A. Khasawneh. "Dynamic State Analysis of a Driven Magnetic Pendulum Using Ordinal Partition Networks and Topological Data Analysis."

23 / 24

Thank you!

Any questions?














Myers, A. D., Chumley, M. M., Khasawneh, F. A., & Munch, E. (2023). Persistent homology of coarse-grained state-space networks. Physical Review E, 107(3), 034303.

24 / 24

Acknowledgements

——————————————————————————————

This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550- 30 22-1-0007

——————————————————————————————

2 / 24
Paused

Help

Keyboard shortcuts

, , Pg Up, k Go to previous slide
, , Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
Esc Back to slideshow