1ssb / Whiplash

A Closed Loop Gradient Descent Algorithm applied to Rosenbrock's function

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Whiplash Gradient Descent and the Analysis of Closed Loop Gradient Descent

Final Paper

Journal Published: https://onlinelibrary.wiley.com/doi/epdf/10.1002/asjc.3153 Please find the detailed paper here: https://arxiv.org/abs/2108.12883.

How to cite

Bhattacharjee, S. S. and Petersen, I. R., Analysis of the Whiplash gradient descent dynamics, Asian J. Control (2023), 1–14, DOI 10.1002/asjc.3153.

Description

This file contains live MATLAB projects and Simulink simulations by: Mr. Subhransu Sekhar Bhattacharjee, U7143478, ANU, under the supervision of Prof. Dr. Ian R. Petersen FAA, College of Engineering and Computer Science, ANU. Please direct any queries regarding the code to Mr. Subhransu Bhattacharjee @ u7143478@anu.edu.au. Please use MATLAB version 2021a for running the .mlx file.

View Whiplash on File Exchange

Our second paper which analyses the Whiplash Gradient Descent for Convex functions has been published in the Asian Journal of Control and was presented at the 13th Asian Control Conference, Jeju Island, Korea. The link to our work is https://arxiv.org/abs/2203.02140. In this paper, we propose a variation on the Landau notation-based complexity theory to prove using numerical computation the convergence rates in continuous time without resorting to rigorous Lyapunov functional analysis. Note the Whiplash Exploration Algorithm has been fully described rigorously in the final paper and I will not be releasing the MATLAB code for it as it seems unnecessary.

About

A Closed Loop Gradient Descent Algorithm applied to Rosenbrock's function

License:GNU General Public License v3.0


Languages

Language:MATLAB 100.0%