GD v.s. CGD for Linear Regression

Gradient Descent (GD) v.s. Conjugate Gradient Descent (CGD) for 2-D Linear Regression

Introduction

This is one part of the course assignment for HKUST-GZ MICS 6000I Physical Design Automation of Digital Systems. This project is alive, maintained by [email protected]. Any discussion or suggestion would be greatly appreciated!

Requirements

• Python 3.9
  • ply 3.11
  • matplotlib 3.5.1
  • logging 0.5.1.2
  • numpy 1.21.5

How to Run

• python main.py
  • dataset: data.txt
  • source code: main.py

Results

• Logs:

  • in GD-CGD.log

• Convergence trajectories:

  • Gradient Descent (GD): in GD-cost-convergence.png
  
  • Conjugate Gradient Descent (CGD): in CGD-cost-convergence.png
  

• 3-D Convergence Animation:

  • Gradient Descent (GD): in GD-result-3d.gif
  
  • Conjugate Gradient Descent (CGD): in CGD-result-3d.gif
  

• Intermediate results (3-D figure) during convergence

  • in result folder