Numerical Optimization#
➡️ Numerical Optimization:
🔻 Objective.
🔸 The aim of this project is to apply different nonlinear numerical optimization methods to solve the Logistic Regression maximum likelihood problem.
🔸 Several numerical algorithms have been programmed, and compared both with each other and with implementations of the Scipy and Sklearn libraries, in terms of computational efficiency and optimality.
🔻 Developments
️ 🔸 Presentation of the optimization problem present in binary logistic regression.
️ 🔸 Implementation of numerical optimization algorithms: Newton, Quasi-Newtons, Gradient Descent and stochastic versions of each.
️ 🔸 Comparative analysis of the implemented methods as well as solvers of the Scipy and Sklearn frameworks.
️ 🔸 Presentation of the logistic regression problem as an optimization problem with constraints.
️ 🔸 Resolution through a numerical optimization method with penalty.