Graph-Clouring In F#

Overview

This project has been conducted in accordance with Massey University directive while undergoing course 159.333 Programming Project.

This project is designed to explore different graph-colouring algorithms and their implementation using F# on the .NET stack. It is a .NET Core 3.1 application.

Requirements

• .NET Core 3.1 SDK.
• VSCode
• Ionide extension for VSCode

The Problem

The Graph Colouring problem is what is known as an NP Complete problem.

Algorithms

Random Colouring Algorithm

The Random Colouring Algorithm shuffles all indices, using the Fisher-Yates shuffle algorithm in this case, to randomise the order in which the vertices are processed.

Psuedo Code

1. Randomly select first vertex and colour it with the lowest chromatic index
2. For each remaining vertex (V-1) do the following:
    a. Choose a vertex at random
    b. Consider the lowest available chromatic index that is not present
        on an adjacent vertex
    c. If all adjacent vertices assign a new chromatic index

Greedy Algorithm

The Greedy Colouring Algorithm doesn't guarantee to use a the minimum number of colours, but it does gaurantee an upper bound on the number of colours - being d+1 where d is the maximum degree of a vertex.

Psuedo Code

1. Colour the first veretex with the lowest chromatic index
2. For each remaining vertex (V-1) do the following:
    a. Consider current vertex and assign the lowest chromatic index that hasn't been used on a previously coloured vertex adjacent to it.
    b. If all used colours appear on adjacent vertices, assign a new chromatic index

Running The Application

To run the application normally:

dotnet clean
dotnet build
dotnet run

To run the application in the debugger:

• Open VSCode
• Set breakpoints
• Open Debug view
• Select Run > Start Debugging (or pressing F5)

Data

The data is set to be taken from ./Data/anonymised.csv at build time.

There is a TODO: in code to implement this as a setting rather than hardcoded.

The data is currently required to follow the proceeding format from the CSV file:

id, exam1, exam2, exam3, exam4,

Any empty exam slots should be indicated by a following comma and don't need to be sequential e.g. 00, AA,, AB,,

NOTE: Whitespaces are trimmed during the csv deserialsation.

NOTE: For formatting of the timetable output at the end it is recommended to stick to 2 character representations of the exams

Settings

The settings are available in ./settings.json and are deserialised when the application runs.

They consist of the following:

{
    "algorithm": "all | greedy | random",
    "rooms": 10
}

NOTE: For formatting of the timetable output at the end it is recommended to stick to a room limit lower than 100.

NOTE: The settings get deserialised as option types, this means that if they are undefined in the JSON file the application will use defaults instead

NOTE: The only algorithms available are "greedy", "random", more were planned to be added.

Settings To Be Implemented

The following are settings that have not yet been implemented because of time constraints:

• "data": data location is hardcoded
• "timeslots": there is currently no restriction on timeslots i.e. number of columns in output timetable

Desired settings:

{
    "algorithm": "all | random | greedy | k-start | lazy-dfs",
    "data": "/location/to/data/file.csv",
    "rooms": 10,
    "timeslots": 10
}