MSK_RES_ERR_INFINITE_BOUND (A numerically huge bound value is specified.)

1

when i run a function, there is an error.
MOSEK Version 9.1.9 (Build date: 2019-11-21 11:34:40)
Copyright © MOSEK ApS, Denmark. WWW: mosek.com
Platform: Windows/64-X86

*** Error(1400): buc[1] is too small
Mosek error: MSK_RES_ERR_INFINITE_BOUND (A numerically huge bound value is specified.)

Status: Error
Optimal value (cvx_optval): NaN

my code is
function [theta]=phase_shift(L,B,T,H_UB,H_IB,H_IE, H_UI,H_UE,h_EJ,h_JB, h_JI,W,l,p,p_m,sigma)
precision = 10^(-2);
r_max = 100;
K =2;
M=10;
for k = 1:K
for j = k:K
v(:,j,k) = diag(H_UI(:,k)’)H_IBW(k,:)’; %10*1
r(j,k) = H_UB(:,k)’*W(k,:)’;
r_m(j,k) = r(j,k)*r(j,k)’ ;
R(:,:,j,k) =[v(:,j,k)*v(:,j,k)’ v(:,j,k)*r(j,k)’; v(:,j,k)‘r(j,k) 0] ;%
end
o(k) = 2^((L - l(k))/(BT));
J(:,:,k) = diag(H_UI(:,k)’)*H_IB;
J_e(:,:,k) = diag(H_UE(:,k)’)*H_IE’;
J_m(:,:,k) = diag(h_JI’)*H_IE;
S(:,:,k) = [J(:,:,k)*J(:,:,k)’ J(:,:,k)H_UB(:,k); H_UB(:,k)'J(:,:,k)’ 0];
S_e(:,:,k) = [J_e(:,:,k)'J_e(:,:,k) J_e(:,:,k)'H_UE(:,k); H_UE(:,k)'J_e(:,:,k) 0];
S_m(:,:,k) = [J_m(:,:,k)J_m(:,:,k)’ J_m(:,:,k)h_EJ; h_EJ’J_m(:,:,k)’ 0];
v_a(:,k) = diag(h_JI’)H_IBW(k,:)’; %101
r_a(k) = h_JB’W(k,:)’;
r_a_m(k) = r_a(k)r_a(k)’;
R_a(:,:,k) = [v_a(:,k)v_a(:,k)’ v_a(:,k)r_a(k); v_a(:,k)'r_a(k) 0];
he_a(k) = H_UE(:,k)'H_UE(:,k);
ha_a(k) = H_UB(:,k)'H_UB(:,k);
end
h_em = h_EJ’h_EJ;
error7=1; r7=1;
E_old = ones(M+1,M+1);
r1_old=[100,100];
d1_old = [0,0];
r2_old=[0,0];
d2_old = [0,0];
while(error7 > precision && r7< r_max)
cvx_begin
cvx_solver mosek
% cvx_solver_settings(‘rmdepconstr’,1)
variable E(M+1,M+1) complex
variables r1(1,K) r2(1,K)
variables d1(1,K) d2(1,K)
expressions c1(1,K)
find E
subject to
for k =1:K
for j = k+1:K
c1(j) = p(j)real(trace(ER(:,:,j,k)))+p(j)(r_m(j,k));
end
p(k)(real(trace(ES_e(:,:,k)))+ he_a(k))10^(20)>=(r1_old(k)-1)d1_old(k)+(r1_old(k)-1)(d1(k)-d1_old(k))+d1_old(k)(r1(k)-r1_old(k));
sum(c1)+sigma+p_m(real(trace(ER_a(:,:,k)))+ r_a_m(k))<=d1(k);
sigma + p_m(real(trace(ES_m(:,:,k)))+ h_em)<=d2(k);
sigma + p_m(real(trace(ES_m(:,:,k)))+ h_em)+p(k)(real(trace(ES_e(:,:,k)))+ he_a(k))>=r2_old(k)d2_old(k)+r2_old(k)(d2(k)-d2_old(k))+d2_old(k)(r2(k)-r2_old(k));
r2_old(k)r1_old(k)+r2_old(k)(r1(k)-r1_old(k))+r1_old(k)(r2(k)-r2_old(k))>=o(k);
end
for m=1:M+1
E(m,m)==1;
end
real(trace(ES(:,:,1)))+ ha_a(1) >= real(trace(ES(:,:,2)))+ ha_a(2);
cvx_end
error7 = norm(E-E_old,inf);
E_old = E;
r1_old=r1;
d1_old = d1;
r2_old=r2;
d2_old = d2;
end
end
where p = [17,38]; l = [93,74]; p_m = 3.5; W = ones(K,N);
R，S,S_e, S_m, R_a, he_a ，there magnitude is 10^(-15)- 10^(-20),

variables r1(1,K) r2(1,K)
variables d1(1,K) d2(1,K)  are auxiliary variable in order to make the problem convex ,  E is a semidefinite matrix.
2

It means somewhere in your model you (or CVX) are trying to add an upper bound that is essentially equivalent to x\leq -\infty, for example x\leq -10^{40} or something of that order.

3

It means I need to rescale the problem?

4

Maybe, or remove huge bounds, or something along these lines. Your code contains constants like 10^{20} and you say some input data are 10^{-20}, that is a huge red warning lamp flashing.