Optimization based on surrogate modeling for analog integrated circuits
Firas Yengui, Lioua Labrak, Patrice Russo, Felipe Frantz, Nacer Abouchi · 2012
The use of surrogate models has become an effective alternative for complex computer simulations. To obtain an optimal result in a reduced optimization time, we present a new approach in which we replace the actual circuit by a surrogate model. First, we introduce and compare results of three different surrogate modeling techniques, such as: kriging, radial based function and rational. We apply each method to circuit performances of the transimpedance amplifier circuit. Second, we compare results of hybrid optimization based on the three different presented surrogate modeling with SPECTRE netlist. The result shows that our approach allows to find optimal solutions in a short time compared to classical approach.