Clustering with an imposed graph information
Iteratively repeat until convergence in Lϵ,α:
(Step i) for a current value of Γ, Eq. 4 is minimized as an unconstrained convex (for example, quadratic) problem wrt. Θ only (this can be done analytically in many cases);
(Step ii) for a current value of Θ, Eq. 4 is minimized to wrt. Γ only, as a constrained convex [for example, as a quadratic programming (QP)] problem.