%% Copyright Radu Dogaru, 1995, 2006, % last update December 8, 2006 % % e-mail: radu_d@ieee.org %------------------------------------------------------- % Appendix to Book: % R. Dogaru: "Universality and Emergent Computation in CNNs, World Scientific 2003", World Scientific, 2003 % http://www.worldscibooks.com/chaos/5053.html %----------------------------------------------------------------------------------- % Returns a matrix ed1 of size N,N corresponding to % up to 13 different possible behaviors of the cell parameter points, % as obtained using the edge of chaos tests in equatiuons (19),(21), page 105 % of the abovementioned book (one-diffusion coefficient, i.e. D1 non-zero). % The program can be easily extended to the case of "two difussion coefficients", i.e. D2 non zero. % using (22) and (24) at page 108. % Here a FitzHugh-Nagumo model with parameters a,b,eps and c is considered % See eq. (3) pp. 99 in book % A parameter point (a,b) is associated with an (i,j) entry of the % ed1 matrix. The value of ed1(i,j) can be as follows: % 1 - cell is passive (P) % 2 - cell is SA (stable and active) in at least one equilibrium point (EQ) % typically there is only one EQ % 3- cell has 2 equilibria that are both SA, and one that is P - unlikely % 4- cell has 3 equilibria that are all SA - unlikely % 5- cell has only one equilibria and it is UA (Unstable ans Active) % 6 - cell has an EQ which is P, an EQ which is SA and the last is UA % in the ablve, the cell can be considered stable and unstable, active and passive ! % 7 - cell has two EQ that are SA and one which is SU % 8 - unlikely % 9 - cell has two EQ that are SU and one which is P % 10 - cell has two EQ that are SU and one which is SA % 11 % 12 % 13 - cell has all three EQ unstable %--------------------------------------------------------------------------- % N is the number of sampples in the interval [amin,amax] and [bmin,bmax] function [ed1]=edge_fhn(a,b,e,c,N) % the first two parameters are given as ranges, % the others are fixed: % For instance [ed1]=edge_fhn([-10 10],[-10 10],-0.1,1,100); vis_edge; % interesting sub-regions within the (a,b) space may be further investigated % using the same program. The second function displays the section within the % (a,b) plane assigning certain colors (defined inside vis_edge.m) to each of the % above mentioned numbers (from 0 to 12) indicating various dynamic qualities % for the cell amin=a(1); amax=a(2); bmin=b(1); bmax=b(2); i=1:N; j=i; a(j)=amin+(amax-amin)*(j-1)/(N-1); b(i)=bmin+(bmax-bmin)*(i-1)/(N-1); ed1=zeros(N,N); for i=1:N for j=1:N [x0]=roots([-b(i), 0, (3*c*b(i)-3), -3*a(j)]); %find the equilibria of FitzHugh Nagumo cell ix=find(imag(x0)==0); x0=x0(ix); % keep only the real valued solutions nsol=size(x0,1); % compute the elements of the Jacobian matrix that are independent % of the number of root FitzHugh-Nagumo model a12=-1; a21=-e; a22=e*b(i); %-------------------------------------------------------------- % The scalars U, T and DEL U=[]; T=[]; DEL=[]; % 1 diffusion case typ=[]; for k=1:nsol % compute the T,DEL and U coefficients required by the local activity tests % and the Jacobian parameter a11 which depends on the particular root a11(k)=c-x0(k)^2; T(k)=a11(k)+a22; U(k)=(a12-a21); DEL(k)=a11(k)*a22-a12*a21; %------------- % test SA, P and SU conditions (19), (21) at p.105 % and assign a number to the quality of the dynamic phenomena if ( (a22<0 & (T(k)<0 & T(k)>a22)) | (a22>0 & T(k)<0) ) & (DEL(k)>0) typ(k)=1; elseif (DEL(k)>0) & (T(k)