École Nationale Supérieure de
Techniques Avancées

Abstract

During this internship, an iterative optimization method called NCL was implemented. For each iteration, an augmented lagrangian like sub problem is solved, through residuals. Under hypothesis, residuals allow to solve unqualified problems. This optimization method was initially developped by MM.\textsc{Orban}, \textsc{Saunders}, \textsc{Judd} and \textsc{Ma}, aiming at solving optimal tax problems. Those are degenerated and with a high number of constraints hence the current difficulties to solve them using usual methods. NCL was implemented in Julia, in order to make it as accessible as possible. Each iteration calls an internal solver, using an interior point method. In contradiction with the state of the art of this method, forcing a warm start initial point in the internal solver is efficient for NCL resolution.