using symmetries and equitable partitions together to find
TRANSCRIPT
![Page 1: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/1.jpg)
Louis Pecora Naval Research Laboratory, Washington, DC
Aaron HagerstromNIST, Boulder, CO
Thomas Murphy IREAP & ECE, University of Maryland, College Park, MarylandRajarshi Roy IREAP & Physics, University of Maryland, College Park, Maryland
Francesco Sorrentino University of New Mexico, New Mexico
Using Symmetries and Equitable Partitions Together to Find All Synchronization
Clusters and Their Stability
SIAM Dynamical Systems Conference May 25, 2017
Joseph Hart IREAP & Physics, University of Maryland, College Park, Maryland
MS168 Advances in Network Synchronization - Part II of II
Per Sebastian Skardal, Jie Sun, and Dane Taylor
Abu Bakar Siddique1 University of New Mexico, New Mexico
![Page 2: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/2.jpg)
What are synchronized clusters?
![Page 3: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/3.jpg)
Cluster Synchronization in complex networks
(various versions in the literature)
Identify the clusters? Are the clusters stable? Complex (large) networks Any dynamics (fixed pt, periodic,
quasiperiodic, chaotic
Identical nodes (oscillators), Identical edges (couplings)Adjacency matrix
Cluster synchronization, an “old” subject (early 2000’s)
dxidt
= F xi( )+σ CijH x j( )j=1
N
∑
![Page 4: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/4.jpg)
![Page 5: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/5.jpg)
Symmetries and clusters in networks
![Page 6: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/6.jpg)
Using symmetry to find clusters.
90o
Network with identical nodes (oscillators) and identical edges
1 0 0 0 00 0 1 0 00 0 0 1 00 0 0 0 10 1 0 0 0
⎛
⎝
⎜⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟⎟
1 23
45
1 0 0 0 00 0 0 1 00 0 1 0 00 1 0 0 00 0 0 0 1
⎛
⎝
⎜⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟⎟
1 0 0 0 00 0 0 0 10 0 0 1 00 0 1 0 00 1 0 0 0
⎛
⎝
⎜⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟⎟
G ={gi}
g1 g2 g3
group
Rg1 Rg2 Rg3
Rgi{ } Representation of the group G g2g1−1 = g3e.g. Rg2Rg1
−1 = Rg3=>
dxidt
= F xi( )+σ CijH x j( )j=1
N
∑ ⇒ RgC = CRg ⇒ C = Rg
−1CRg
![Page 7: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/7.jpg)
Random Graphs
![Page 8: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/8.jpg)
Random networks 11 nodes – 9 random edges
0 symmetries
G.gens()= [(7,10), (6,7), (5,6), (4,8), (2,4)(8,9), (1,5), (1,11)]
8640 symmetries
{gi}
![Page 9: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/9.jpg)
Computational Group Theory
Groups, Algorithms, Programming - a System for Computational Discrete Algebra
GAP
Sagehttp://www.sagemath.org/
sage: x = var('x')sage: solve(x^2 + 3*x + 2, x)[x == -2, x == -1]
Pythonfor i,row in enumerate(Adjmat):
rsum= row.sum()Cplmat[i,i]= -rsum
print Cplmat
Free ! (open source)
![Page 10: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/10.jpg)
G.order(), G.gens()= 8640 [(9,10), (7,8), (6,9), (4,6), (3,7), (2,4), (2,11), (1,5)]
node sync vectors: Node 2
orb= [1, 5]nodeSyncvec [0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0]cycleSyncvec [1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0] Node 1
orb= [2, 4, 11, 6, 9, 10]nodeSyncvec [1, 0, 1, 0, 1, 0, 1, 0, 0, 1, 1]cycleSyncvec [0, 1, 0, 1, 0, 1, 0, 0, 1, 1, 1] Node 4
orb= [3, 7, 8]nodeSyncvec [0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0]cycleSyncvec [0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0]
-7. 1. 0. 1. 1. 1. 0. 0. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. -6. 1. 0. 1. 0. 0. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. -7. 1. 0. 0. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 0. 1. 0. 1. 0. 1. -6. 0. 1. 1. 1. 0. 1. 0. 1. 0. 1. 0. -6. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10.
Stability: Coupling Matrix
d δxidt
= DF xi( )δxi +σ CijDH x j( )j=1
N
∑ δx j
δxi →δxi + b ∀i ∈Cluster
M. Golubitsky, I. Stewart, and D.G. Schaeffer, Singularities and groups in bifurcation theory, Vols. I & II (Springer-Verlag, New York, NY, 1985).
(1) No computer code for large networks (2) Did not examine stability of clusters.
![Page 11: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/11.jpg)
-6.00 -3.46 0.0 -3.46 -5.00 -4.24 0.0 -4.24 -6.00 -8.00 -6.00 -6.00 -11.00 -11.00 -11.00 -11.00 -11.00
Mg=1 Trivial Representation
Synchronization Manifold
Transverse Manifold (sub-blocks associated with each cluster)
All other Representations
-7. 1. 0. 1. 1. 1. 0. 0. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. -6. 1. 0. 1. 0. 0. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 0. 1. -7. 1. 0. 0. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 0. 1. 0. 1. 0. 1. -6. 0. 1. 1. 1. 0. 1. 0. 1. 0. 1. 0. -6. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. -10.
dξλ
dt= DF s( )+ ΛλDH s( )( ) ⋅ ξλ
d δxidt
= DF xi( )δxi +σ CijDH x j( )j=1
N
∑ δx j
Simplifying the variational matrix
G= group of all symmetries
∀g∈ Irreducible Representations(IRR)
∃ T0
0TRgT
−1 =
⎛
⎝
⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟
g∀Rg{ } It is reducible if
Rg{ } is a representation of G
The software does not supply T
T block diagonalizes C C = R
g
−1CRg
![Page 12: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/12.jpg)
Experimental confirmation of symmetry clusters
Aaron HagerstromRajarashi Roy
Thomas MurphyFrancesco Sorrentino
![Page 13: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/13.jpg)
!How to implement maps in experiment!
4
Phase written to SLM Intensity measured by camera
Adjacency matrix Add phase shift to destabilize fixed point φ=0
Self feedback strength
Coupling strength
Spatial Light Modulator
![Page 14: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/14.jpg)
11 node random network (video)32 Symmetries4 nontrivial clusters1 trivial cluster5 x 5 Sync block
β
![Page 15: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/15.jpg)
![Page 16: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/16.jpg)
![Page 17: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/17.jpg)
Chimera in identical synchronized network (Not just for phase oscillators)
Chimeras, Cluster States, and Symmetries: Experiments on the Smallest Chimera - Joseph Hart
MS 122
Not transients
![Page 18: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/18.jpg)
Constructing non-symmetry clusters
Beyond symmetry clusters
![Page 19: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/19.jpg)
Forming Equitable and Laplacian Clusters from Symmetry (orbital) Clusters
Def: An Equitable Partition of a graph is a collection of disjoint subsets Ci ( Ci ∩C j =∅ ) of all the nodes such that, if node v∈Ci has d connections to nodes in C j then for any other node u ∈Ci has d connections to nodes in C j .
C1C2C3
Besides being an Equitable Partition (EP) this is also an Orbital Partition (OP) from the symmetries.
Every OP is an EP, but the converse is not true. There are EPs that are not OPs.
C1C2
Example: McKay’s graph
An algorithm from Igor Belykh* provides an efficient method for finding the coarsest EP. However finding finer EPs is a hard problem. This is important since there may be finer EPs that are still not OPs. An example coming up.
* I. Belykh and M. Hasler, CHAOS 21, 016106 (2011)
Problem 1: Find finer EPs that are not OPs.
Similar to groupoid synchronization criterion of Golubitsky, Stewart and Török (SIAM J. APPLIED DYNAMICAL SYSTEMS Vol. 4, No. 1, pp. 78–100 (2005)
![Page 20: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/20.jpg)
Problem 2: Stability calculations. Reducing the transverse manifold in the variational equations.
d δxidt
= DF xi( )δxi +σ CijDH x j( )j=1
N
∑ δx j
OPs (Symmetry Clusters)
EPs (Equitable Clusters)
Ref: Schaub et al.*
each block is associated with particular clusters. Master Stability Function for each cluster.
• Stability calculations scale badly (N2)
• Stability of individual clusters is unknown.
• Type of desynchronization bifurcation is unknown.
1 1 0 1 0 0 0 1 0 00 0 1 0 1 0 0 0 0 10 0 0 0 0 1 1 0 1 0
⎛
⎝
⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟
VIs it even possible to block diagonalize the transverse manifold?
dξλ
dt= DF s( )+ ΛλDH s( )( ) ⋅ ξλ
*M. Schaub, et al., Chaos 26 , 094821 (2016)
Make synchronization manifold basis and find the orthogonal compliment (the transverse space, V , this will separate out the synchronization manifold, but will not block diagonalize the transverse manifold.
0
0CiCi IRR
0
0
![Page 21: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/21.jpg)
Constructing an EP from an OP The Abu network.
Symmetries->OP orbitals= [1,2,3,4], [5,7,8,10], [6,9]
1 2
3 4
56
7
89 1
0
Make the EP by merging clusters from its OP refinement
[5,7,8,10] U [6,9]
Observation: We can always construct an EP from an OP refinement. There is always a subgroup of the original symmetry group that will be a
refinement of any EP
EP ->Equitable Clusters= [1,2,3,4], [5,7,8,10,6,9]
![Page 22: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/22.jpg)
Construction of a block diagonalized variational matrix for the EP
-4.00 -1.41 -1.41 -1.41 -3.00 2.00 -1.41 2.00 -4.00 -5.00 -3.00 -4.00 -1.41 -1.41 -1.41 -5.00 2.00 -1.41 2.00 -4.00 -4.00 -4.00 T
00 (0,1,0,1,1,0,1,0,0,0)
(1,0,0,0,0,1,0,0,0,0)
(1,1,0,1,1,1,1,0,0,0)merged clusters
s=
u1=Ts
v2
v1
u1u2
new sync direction
new transverse direction
T
for the new EP -4.00 -2.45 -2.45 -2.00 -5.00 -5.00 -3.00 -4.00 -1.41 -1.41 -1.41 -5.00 2.00 -1.41 2.00 -4.00 -4.00 -4.00
00v1
v2
v3
![Page 23: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/23.jpg)
Laplacian coupling - diag = – row sum
12
3 4
5−3 1 0 1 11 −3 1 0 10 1 −3 1 11 0 1 −3 11 1 1 1 −4
⎛
⎝
⎜⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟⎟
dxidt
= F xi( )+σ CijH x j( )j=1
N
∑
Non-symmetry, Non-equitable clusters
0Complete cluster xi = s ∀i
dsdt
= F s( )+σ CijH s( )j=1
N
∑ = F s( )+σH s( ) Cijj=1
N
∑ = F s( )
Combine clustersxi = s i = 1,3,5
12
3 4
5dsdt
= F s( )+σ H x2( )+ H x4( )+ H s( )− 3H s( )⎡⎣ ⎤⎦
= F s( )+σ H x2( )+ H x4( )− 2H s( )⎡⎣ ⎤⎦
![Page 24: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/24.jpg)
Hierarchy of Adjacency and Laplacian Clustering
Adjacency and
Laplacian
Adjacency and
LaplacianLaplacian
Orbital Partitions Symmetry Clusters
Equitable Partitions
[Symmetry Clusters]
External Equitable Partitions
[Symmetry Clusters]
…and simplify the variational equation.
BuildingBlocks
![Page 25: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/25.jpg)
![Page 26: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/26.jpg)
Conclusions and remarks
Weighted edges/couplings
Different oscillators
Encompasses or overlaps other "phenomena"
Cluster sync, Partial sync, Remote sync, Some Chimera states
We can analyze all potential cluster formations in Adjacency and Laplacian networks
Bifurcation forms
Normal forms & symmetry
M. Golubitsky, I. Stewart, and D.G. Schaeffer, Singularities and groups in bifurcation theory, Vols. I & II (Springer-Verlag, New York, NY, 1985).
![Page 27: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/27.jpg)
1. Cluster Synchronization and Isolated Desynchronization in Complex Networks with Symmetries, Pecora, Sorrentino, Hagerstrom, Murphy, and Roy, Nature Communications, 5, 4079 (13 June 2014)
B.D. MacArthur and R.J. Sanchez-Garcia, Spectral characteristics of network redundancy,"Physical Review E 80, 026117 (2009).B.D. MacArthur, R.J. Sanchez-Garcia, and J.W. Anderson, \On automorphism groups of networks,"Discrete Appl. Math. 156, 3525 (2008).
Thanks to:
2. Complete Characterization of Stability of cluster Synchronization in Complex Dynamical Systems, Sorrentino, Pecora, Hagerstrom, Murphy, and Roy, Science Advances 5, 011005–1–17 (2015).
3. Francesco Sorrentino and Louis Pecora, Approximate cluster synchronization in networks with symmetries and parameter mismatches," CHAOS 26, 094823 (2016);
4. Joseph D. Hart, Kanika Bansal, Thomas E. Murphy, and Rajarshi Roy, Experimental observation of chimera and cluster states in a minimal globally coupled network,CHAOS 26, 094801 (2016)
Hagerstrom, A. Network Symmetries and Synchronization (https://sourceforge.net/projects/networksym/, 2014).
Software
![Page 28: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/28.jpg)
Questions?
![Page 29: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/29.jpg)
Learning Group Theory and Group Representation Theory
M. Tinkham, Group Theory and Quantum Mechanics (McGraw-Hill, New York, NY, 1964)representation theory in the first three chapters (~ 37 pages).
M. Golubitsky, I. Stewart, and D.G. Schaeffer, Singularities and groups in bifurcation theory, Vols. I & II (Springer-Verlag, New York, NY, 1985).
online: http://www.learnerstv.com/Free-Maths-Video-lectures-ltv759-Page1.htm Learner's TV
Nadir Jeevanjee, An Introduction to Tensors and Group Theory for Physicists, Birkhäuser (New York, NY, 2010)
Michael T. Vaughn, Introduction to Mathematical Physics, Chapter 9 (Wiley-VCH)
![Page 30: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/30.jpg)
B.D. MacArthur, R.J. Sanchez-Garcia, and J.W. Anderson, \On automorphism groups of networks," Discrete Appl. Math. 156, 3525 (2008).
Geometric decomposition into subgroups.
Number of Symmetries
> 88% of nodes are in clusters in all above networks
Number of Edges
Number of Nodes
![Page 31: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/31.jpg)
Other networks
![Page 32: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/32.jpg)
Symmetries and clusters in networks with different topologies
Random Scale-free Tree
Random Bipartite
10,000 realizations of each type
1060 symmetries
A.-L. Barabasi and R. Albert, \Emergence of scaling in random networks," Science 286, 509{512(1999).
N= 100 nodes (oscillators)
ndelete= 50
Sage routine RandomBipartite().
10 nodes
90 nodes
![Page 33: Using Symmetries and Equitable Partitions Together to Find](https://reader031.vdocuments.site/reader031/viewer/2022020623/61f02273b3b830775079d485/html5/thumbnails/33.jpg)
Electric power grid of Nepal N=15
Cluster synchronization?
86,400 symmetries15 Nodes3 clusters2 trivial clusters3 subgroups