Parameter dimensioning algorithms of the PNNI complex node model with bypasses

In this paper, we propose optimization approaches for the parameter determi nation of the PNNI complex node model. Two optimal objectives are discussed : Least Square Approximation and Maximum Deviation Minimization. For each o bjective, wt propose two practical criteria for setting up by passes: Maxim um Difference Removal and Largest Deviation Removal. Generalized inverse of matrix and linear programming techniques are used to find the solutions. T he numerical results show that the least square approximation with the larg est deviation removal criteria has the best performance as the number of by passes increases.