problem with a for cycle

I have to iterate a recursive system of difference equation where in the matrix of coefficient the diagonal depends on the previous time vector. The code run if into the for cylce i use mu_0 (the initial vector) but i'm not able to make use of the mu_(t-1) vector (the vector of the previous period).

Here the code. If I didn't explain well let me know

% initialize vector 
mu_0 = rand([n, 1]);

% matrice di adiacenza
A = rand(n, n);

   % normalizzo la matrice
   A= diag(1./sum(A,2))*A;

    % definisco la diagonale
   da=diag(diag(A));

   % separo la matrice dalla diagonale, A1 è il netowrk senza pesi
   % individuali
   A1=A-diag(diag(A));

for t=1:100

     % bayesian updating:

     % likelihood
     p_t=1/2+ rand(n, 1)/4;

     p1_t=diag(p_t);

%%PROBLEM
   mu_t=mu_t-1;

     m_t=p1_t*mu_t+(1-p1_t)*(1-mu_t)  ;

     % eta è l'updating bayesiano senza il prior (mu)

     eta_t=p_t./m_t;
      da1=da*eta_t;

mu_t = (A1^t+diag(da1))*mu_t-1;

%or mu_t = (A1^t+diag(da1))^0*mu_t;

end