540 likes | 688 Views
UK-Japan Winter School Dynamics and Complexity. Dynamics of Networks 2 Synchrony & Balanced Colourings. Ian Stewart Mathematics Institute University of Warwick. Network. A network or directed graph consists of a set of: • nodes or vertices or cells connected by
E N D
UK-Japan Winter School Dynamics and Complexity Dynamics of Networks 2Synchrony & Balanced Colourings Ian Stewart Mathematics Institute University of Warwick
Network A network or directed graph consists of a set of: • nodes or vertices or cells connected by • directededges or arrows
Admissible ODEs dx1/dt = f(x1,x2,x4,x5) dx2/dt = f(x2,x1,x3,x5) dx3/dt = g(x3,x1,x4) dx4/dt = g(x4,x2,x4) dx5/dt = h(x5,x4)
Pattern of Synchrony Cells c, d are synchronous on some trajectory x(t) if xc(t) = xd(t) for all t Defines an equivalence relation ~ for which c ~ d if and only if c, d are synchronous Call ~ a pattern of synchrony
Colouring Interpretation More intuitively, colour cells with different colours, so that cells have the same colour if and only if they are synchronous Synchrony Space or Polydiagonal D~ = {x : xc = xd whenever c ~ d} This is the set of all cell states with the pattern of synchrony ~ It forms a vector subspace of phase space P
dx1/dt = f(x1,x2,x4,x5) dx2/dt = f(x2,x1,x3,x5) dx3/dt = g(x3,x1,x4) dx4/dt = g(x4,x2,x4) dx5/dt = h(x5,x4)
dx1/dt = f(x1,x1,x3,x5) dx1/dt = f(x2,x1,x3,x5) dx3/dt = g(x3,x1,x3) dx3/dt = g(x3,x1,x3) dx5/dt = h(x5,x3) x1=x2 x3=x4
dx1/dt = f(x1,x1,x3,x5) dx3/dt = g(x3,x1,x3) dx5/dt = h(x5,x3) restricted equations
dx1/dt = f(x1,x1,x3,x5) dx3/dt = g(x3,x1,x3) dx5/dt = h(x5,x3) restricted to synchrony space {(x1,x1,x3,x3,x5)}
dx1/dt = f(x1,x1,x3,x5) dx3/dt = g(x3,x1,x3) dx5/dt = h(x5,x3) quotient network restricted to synchrony space {(x1,x1,x3,x3,x5)}
States on the synchrony space satisfy the restricted equations and have the pattern of synchrony determined by the colouring The dynamics of these states is determined by the quotient network
Robust Synchrony A pattern of synchrony ~ is robust if D~ is invariant under all admissible vector fields That is: any initial condition with that pattern continues to have that pattern for all time
Balanced Patterns and Colourings A pattern of synchrony ~ (or the associated colouring) is balanced if cells with the same colour have input sets with the same colours (of tail cells) up to some type-preserving permutation (input isomorphism) Theorem A pattern of synchrony ~ is robust if and only if it is balanced
Balanced colouring: cells with same colour have input sets with same colours
Example—the Square Lattice Nearest-neighbour coupling Balanced 2-colouring
Example—the Square Lattice For this pattern, interchanging black and white along any upward sloping diagonal leads to another balanced 2-colouring
The Origin of Species Allopatric speciation Geographical or other barrier prevents gene-flow Easily understood and WYSIWYG
The Origin of Species Sympatric speciation No such barrier; species split in the same location even with panmixis Sympatricspeciation appears paradoxical, but can be viewed as a form of symmetry-breaking
Darwin’s Finches Galápagos Islands
Darwin’s Finches 1. Large cactus finch (Geospiza conirostris) 2. Large ground finch (Geospiza magnirostris) 3. Medium ground finch (Geospiza fortis) 4. Cactus finch (Geospiza scandens) 5. Sharp-beaked ground finch (Geospiza difficilis) 6. Small ground finch (Geospiza fuliginosa) 7. Woodpecker finch (Cactospiza pallida) 8. Vegetarian tree finch (Platyspiza crassirostris) 9. Medium tree finch (Camarhynchus pauper) 10. Large tree finch (Camarhynchus psittacula) 11. Small tree finch (Camarhynchus parvulus) 12. Warbler finch (Certhidia olivacea) 13. Mangrove finch (Cactospiza heliobates)
Darwin’s Finches 1. Large cactus finch (Geospiza conirostris) 2. Large ground finch (Geospiza magnirostris) 3. Medium ground finch (Geospiza fortis) 4. Cactus finch (Geospiza scandens) 5. Sharp-beaked ground finch (Geospiza difficilis) 6. Small ground finch (Geospiza fuliginosa) 7. Woodpecker finch (Cactospiza pallida) 8. Vegetarian tree finch (Platyspiza crassirostris) 9. Medium tree finch (Camarhynchus pauper) 10. Large tree finch (Camarhynchus psittacula) 11. Small tree finch (Camarhynchus parvulus) 12. Warbler finch (Certhidia olivacea) 13. Mangrove finch (Cactospiza heliobates)
The Origin of Species Interaction network — all-to-all coupled in the panmictic case
The Origin of Species With generic nonlinear dynamics, instability of the single-species state leads to symmetry-breaking bifurcation. Universal features are: Split is a jump bifurcation Mean phenotypes stay constant Usual split is to two species
The Origin of Species Simulation with 50 nodes
Darwin’s Finches Geospiza fuliginosa Geospiza fortis Albemarle/ Isabela Daphne Crossman/ Los Hermanos Character displacement Not speciation, but might be OK as a surrogate Beak size in millimetres
J.Cohen and I.Stewart. Polymorphism viewed as phenotypic symmetry-breaking, in Nonlinear Phenomena in Biological and Physical Sciences (eds. S.K.Malik, M.K.Chandra-sekharan, and N.Pradhan), Indian National Science Academy, New Delhi 2000, 1-63 I.Stewart. Self-organization in evolution: a mathematical perspective, Nobel Symposium Proceedings, Phil. Trans. Roy. Soc. Lond. A361 (2003) 1101-1123. I.Stewart., T.Elmhirst and J.Cohen. Symmetry-breaking as an origin of species, in Bifurcations, Symmetry, and Patterns (eds. J. Buescu, S. Castro, A.P.S. Dias, and I. Labouriau), Birkhäuser, Basel 2003, 3-54. I.Stewart. Speciation: a case study in symmetric bifurcation theory, Univ. Iagellonicae Acta Math.41 (2003) 67-88.
Rigid Equilibria Equilibrium States x*is an equilibrium state for an admissible f if and only if f(x*) = 0 Associated Pattern of Synchrony Use the cell coordinates of x* to define a colouring: then c and d have the same colour if and only if x*c = x*d Determines an equivalence relation ~x*
Hyperbolicity An equilibrium x* is hyperbolic if the Jacobian Df|x* has no eigenvalues on the imaginary axis (and in particular is nonsingular). Theorem If x* is hyperbolic and g = f+ep is a small admissible perturbation of f, then g has a unique zero y* near x*.
Rigidity Let x*be a hyperbolic equilibrium state for an admissible f. Then the pattern of synchrony ~x* is rigid if it is the same as ~y* for the nearby equilibrium y* of any sufficiently small perturbation g. More carefully: if x*is a hyperbolic equilibrium state for an admissible f, we can define its rigid pattern of synchrony to be ~x*rig, where c ~x*rig d if and only if yc = yd for every nearby equilibrium y* of any sufficiently small perturbation g. Then ~x*rig is rigid.
Rigidity Theorem [Rigid implies balanced] Let x* be a hyperbolic equilibrium state for an admissible f. Then ~x*rig is balanced. The proof is nontrivial: see Golubitsky, Stewart, and Török. Stronger Result—Transversality Aldis (unpublished) has proved a similar theorem with ‘hyperbolic’ replaced by ‘transverse’, which means that the Jacobian Df|x* has no zero eigenvalues (is nonsingular). This is stronger because Hopf bifurcation points (nonzero imaginary eigenvalues) are permitted.
Rigid Periodic States Is there an analogous theorem for periodic states? A periodic state x(t) is hyperbolic if its linearized Poincaré map (Floquet map) has no eigenvalues on the unit circle. A pattern of synchrony, or of phase shifts, is rigid if it persists after any sufficiently small admissible perturbation.
Rigid Synchrony Conjecture Conjecture If x(t) is a hyperbolic periodic state, then rigidly synchronous cells have synchronous input cells (up to some input isomorphism). Theorem If the RSC is true, and x(t) is a hyperbolic periodic state, then the relation of rigid synchrony is balanced.
Rigid Phase Conjecture Conjecture If x(t) is a hyperbolic periodic state, then rigidly phase-related cells have phase-related input cells (up to some input isomorphism) with the same phase relations. Theorem If the RPC is true, then the quotient network (for the relation of rigid synchrony) has a cyclic group of global symmetries, and the phase relations are among those for a dynamical system with this symmetry .
UK-Japan Winter School Dynamics and Complexity Dynamics of Networksto be continued... Ian Stewart Mathematics Institute University of Warwick
Network Each cell has a cell-type and each arrow has an arrow-type, allowing us to require the cells or arrows concerned to have ‘the same’ structure. In effect these are labels on the cells and arrows. Abstractly they are specified by equivalence relations on the set of cells and the set of arrows.
Network Arrows may form loops (same head and tail), and there may be multiple arrows (connecting the same pair of cells). Special case: regular homogeneous networks. These have one type of cell, one type of arrow, and the number of arrows entering each cell is the same. This number is the valency of the network.
Regular Homogeneous Network This is a regular homogeneous network of valency 3.
Network Enumeration Number of topologically distinct regular homogeneous networks on N cells with valency v
Network Dynamics To any network we associate a class of admissible vector fields, defining admissible ODEs, which consists of those vector fields F(x) That respect the network structure, and the corresponding ODEs dx/dt = F(x) What does ‘respect the network structure’ mean?
Admissible ODEs The input setI(c) of a cell c is the set of all arrows whose head is c. This includes multiple arrows and loops.
Admissible ODEs Choose coordinates xc Rkfor each cell c. (We use Rk for simplicity, and because we consider only local bifurcation). Then dxc/dt = fc(xc,xT(I (c))) where T(I(c)) is the tuple of tail cells of I(c).
Admissible ODEs dxc/dt = fc(xc,xT(I (c))) dx1/dt = fc(x1,x1, x2, x3, x3, x4, x5, x5, x5)
Admissible ODEs for the example network: dx1/dt = f(x1, x2,x2,x3) dx2/dt = f(x2, x3,x4,x5) dx3/dt = f(x3, x1,x3,x4) dx4/dt = f(x4, x2,x3,x5) dx5/dt = f(x5, x2,x4,x4) Where f satisfies the symmetry condition f(x,u,v,w) is symmetric in u, v, w
dx1/dt = f1(x1,x2,x4,x5) dx2/dt = f2(x2,x1,x3,x5) dx3/dt = f3(x3,x1,x4) dx4/dt = f4(x4,x2,x4) dx5/dt = f5(x5,x4) domain condition