Global optimization of neural network-based electrothermal model for GaN transistors

This paper presents an efficient artificial neural network (ANN) electrothermal modeling approach applied to GaN devices. Global optimization based learning procedures, including genetic algorithm (GA), Multi-Verse (MV) and Moth-Flame (MF) have been implemented and validated. The three modeling approaches show a very good fitting for actual measurement with competitive results for recent MV and MF techniques with respect to the widely used GA one.