Open AccessOn Gradient Descent and Co-ordinate Descent methods and its variants.
Author(s) -
Sajjadul Bari,
Md. Rajib Arefin,
Sadia Jahan
Publication year - 2020
Publication title -
the aiub journal of science and engineering
Language(s) - English
Resource type - Journals
eISSN - 2520-4890
pISSN - 1608-3679
DOI - 10.53799/ajse.v19i3.103
Subject(s) - descent (aeronautics) , coordinate descent , gradient descent , minification , gradient method , mathematical optimization , computer science , optimization problem , nonlinear conjugate gradient method , mathematics , algorithm , artificial intelligence , engineering , artificial neural network , aerospace engineering
This research is focused on Unconstrained Optimization problems. Among a number of methods that can be used to solve Unconstrained Optimization problems we have worked on Gradient and Coordinate Descent methods. Step size plays an important role for optimization. Here we have performed numerical experiment with Gradient and Coordinate Descent method for several step size choices. Comparison between different variants of Gradient and Coordinate Descent methods and their efficiency are demonstrated by implementing in loss functions minimization problem.