A Fredholm integral equation of the second kind with separable integral kernel is may be solved as
$y(x)=f(x)+ \displaystyle\int_a^b K(x,t)y(t)dt$
then,
$\displaystyle\int_a^b y(x)N_i(x)dx = \displaystyle\int_a^b f(x)N_i(x)dx + \lambda \displaystyle\sum_{j=1}^nc_j \displaystyle\int_a^b M_j(x)N_i(x)dx$
and the idea is to solve the system of equations.
but if I find a system of the style $\displaystyle\int_a^b y(x)N_i(x)dx - \displaystyle\int_a^b f(x)N_i(x)dx - \lambda_1 \displaystyle\sum_{j=1}^nc_j \displaystyle\int_a^b M_j(x)N_i(x)dx = \displaystyle\int_a^b y(x)K_i(x)dx - \displaystyle\int_a^b f(x)K_i(x)dx - \lambda_2 \displaystyle\sum_{j=1}^nc_j \displaystyle\int_a^b H_j(x)K_i(x)dx$ It shouldn't have a solution because the number of unknowns doubles, right?
Actually I have not been able to write in matrix terms, and I think that it is not possible. I could ask for $K$ and $N$ to be orthogonal, but that doesn't help along the way either.
thanks for your help