COMP3932: Synoptic Project Repository

About The Project

This project is an investigation into a technique for clustering that uses the Graph Laplacian operator: Spectral Clustering. Implemented is a modular, customisable, packaged pipeline for this machine learning technique, a benchmarking framework that gathers results on runtime and performance, and a set of notebooks that demonstrate initial research and analysis into how this technique works, and its comparison to alternative clustering algorithms.

Getting Started

To get a local copy up and running follow these simple example steps.

Prerequisites

To use the package and run benchmarking you will require the following prerequisites:

• Python 3.X
• Installation of Make, such that you can run Makefile commands

Installation

To install the custom python packaged solution to your local package manager, follow these instructions:

1. Clone the repo

   git clone https://github.com/uol-feps-soc-comp3932-2324-classroom/synoptic-project-cassimpatel.git;

2. Create a virtual environment

    python3 -m venv venv;
    source venv/bin/activate;

3. Install requirements (including source package)

   pip install -r requirements.txt;

Usage

To use the package itself, you may import it like any other Python package:

from src.SpectralClustering    import SpectralClustering
from src.data_generation       import make_many_moons
from src.pipeline_transformers import (
    NullTransformer,
    affinity       ,
    refinement     ,
    laplacian      ,
    decomposition  ,
    embedding      ,
    clustering     ,
    confidence     ,
)

See examples of this usage within the analysis notebooks in prelim_analysis/. For details on parameters provided, see the source code.

NOTE: when running these notebooks, make sure you use the virtual environment you created earlier, so the installation of the source file can be found.

Testing & Benchmarking

To run the full testing framework use the following command:

make test;

This command will gather all tests, run them, and output results to a log file you will find in tests/results/ as tests complete.

Contact

Cassim Patel - 201394797 - [email protected]