Algorithm 4:

Estimating model parameters required to compute ensemble weights using SUEL.MF

Input:	The assignments ($C_k,k\in \left\{1,\dots ,K\right\}$), prediction scores ($f_i\left(x_t\right),i\in \left\{1,\dots ,M\right\},t\in \left\{1,\dots ,N\right\}$), number of latent groups $(K),B,U,V$.
1.	Compute the elements of $\mathit{\text{b}}$ from $B$.
2.	Compute $\sqrt{\frac{1-\pi }{\pi }}{\mu }_k^0\text{s}$ as the eigenvectors corresponding to the leading eigenvalue of $UV{U}^T$.
3.	Estimate the accuracy of each predictive score as $w_i=-\frac{b_i{\mu }_k^0}{\pi -(1-\pi )b_i^2{\mu }_k^{0^2}}$.
Output:	Compute the ensembled scores as $g\left(x_t\right)={\sum }_{i=1}^M{w}_i{f}_i\left(x_t\right)$.