README

Introduction

This is a final project that I did as a student for course ORI 391.Q nonlinear programming that I took in Spring 2024 with Professor Bollapragada. The assignment involved implementing several different optimization algorithms on several different problems and comparing success minimization metrics and convergence properties.

Browse the Python+PyTorch source code here. Generated results can be seen here

Instructions to Run Code

To run this code, first setup and active a Python virtual environment. All the other instructions in this README will assume that such an environment has been setup and activated.

virtualenv setup

cd nonlinear-programming
rm -rf venv
# This is assuming you are using Homebrew. If you are using some other environment, adjust this command accordingly.
/opt/homebrew/opt/python@3.12/bin/python3.12 -m venv venv
source ./venv/bin/activate
pip3 install --upgrade pip setuptools
pip3 install -r requirements.txt

Run code lint and unit tests

Run flake8 lint code verification and pytest unit tests:

make flake8
make test

See the function that matches the final project assignment

The final project assignment specifies a function of the form:

[x, f] = optSolver-TeamName(problem,method,options)

That function exists in final_project.py with some example usage.

Generate tabular results in .html format

# optionally remove any old results.
rm -rf generated/

python generate_html_benchmarks.py

# open .html files in generated/ directory.

Generate line search parameter comparison results in .csv format

# optionally remove any old results.
rm -rf generated/

python evaluate_line_search_parameters.py