Data fitting with signomial programming compatible difference of convex functions

Abstract Signomial Programming (SP) has proven to be a powerful tool for engineering design optimization, striking balance between the computational efficiency of Geometric (GP) and extensibility more general methods optimization. While techniques exist fitting GP compatible models data, no have been proposed that take advantage increased modeling flexibility available in SP. Here, new Difference Softmax Affine function is constructed by utilizing existing Convex (DC) functions. This class fit data log–log space becomes either signomial or set signomials upon inverse transformation. Examples presented here include simple test cases 1D 2D, performance NACA 24xx family airfoils. In each case, RMS error driven less than 1%.