Arthur Robertson
A Look Back on Coursework
Year 1 Coursework: Reviewed.
[ coursework , python , c , java ]
01 July 2023

Introduction

As my first year of university comes to a close, I thought it would be fun to look back on some of the coursework I’ve completed. Unfortunately, I’m not allowed to share the code for any of these projects (university rules), but I can share my thoughts.

Saturated Reverse Polish Notation Calculator

Our first introduction to coursework at University was to create a Satured Reverse Polish Notation (SRPN) Calculator. Reverse Polish Notation is a mathematical notation where operators follow their operands. For example, 1 + 2 would be written as 1 2 +. It is used in many computer systems, as it is easier to parse than infix notation. Rather than being given a set of requirements and documentation, our job was to reverse engineer the compiled program and implement as many of the features we could find in Python. This proved quite challenging, as there was a whole host of features to be discovered by trial and error, with limited time.

The project required the use of a stack to store the numbers and allow them to be evaluated in postfix notation, instead of the infix notation we are all used to. This was a great introduction to the concept of stacks, and how they can be used to solve problems.

Overall I enjoyed the project and thought it was a good start to the year.

Ratings

Sudoku Solver

The next project as part of the Artificial Intelligence module was to build a Sudoku Solver in Python. This was quite a step up from the previous project and required applying several artificial intelligence concepts we had been taught. The project criteria were simple: given a sudoku puzzle, return the solved puzzle, with computation time taken to solve being a factor in the marking.

First of all, we were encouraged to implement a basic backtracking algorithm using recursion. This gave us an idea of how to approach the task but were too slow to get a good grade. Constraint Propogation was the next step and allowed us to drastically reduce the number of possibilities for each cell and speed up the algorithm. Finally, I implemented a search algorithm to find the cell with the least number of possibilities and try each of them in turn. This proved sufficient to get the top marks.

Ratings

Dungeon of Doom

The third project of the year was to build a 2D ASCII dungeon crawler game in Java. We were given a hefty specification with a set of requirements, involving a player, items, a map, and an enemy bot. The project was split into two parts, with the first part being to implement the game logic, and the second part being to implement the enemy bot.

The trickiest part of the project was the enemy bot, which required the implementation of a pathfinding algorithm. This involved a lot of trial and error and debugging, but eventually, I managed to get it working perfectly. The bot would find the shortest path to the player, and follow it, avoiding obstacles along the way. This was my first introduction to Java and I found it quite enjoyable.

Ratings

C and Java Exercises

The C and Java Exercises were two separate bits of coursework, but there’s not much to say about them. They were a series of small exercises to get us used to the syntax of the languages and were quite enjoyable. I found the C exercises more challenging, as I had never used C before, but I found it quite enjoyable to learn a new language. The problems were similar to the likes of Leetcode, with a difficulty of maybe Easy to Medium. A question implementing an XOR linked list was quite fun, and there was also a question involving multithreading file handling in Java, which was interesting.

Ratings

Spam or Ham Classifier

Now onto the final and favourite project of the year. The Spam or Ham Classifier was a project to build a spam classifier using whatever implementation we wanted, as long as it was in Python and only used NumPy. A classifier is a machine learning algorithm that takes a set of features and predicts a class. In this case, the classes were spam or ham, representing a spam email or a non-spam email. In a traditional machine learning project, we would have to extract the features from the emails ourselves, but this was already done for us. A feature could be the number of times a certain word appears in the email, or the length of the email, for example, but in our case, it was anonymised, so we didn’t know what the features actually represented.

We were given a dataset of 1,000 emails with extracted features, and our goal was to build a classifier to be marked on an unseen dataset of emails. Over the year we learned several different techniques for classification, including Naive Bayes and Nearest Neighbour, but we were encouraged to use whatever we wanted.

I experimented extensively on the dataset with different techniques. I eventually settled on a voting algorithm which combined the results of several different classifiers. This included Nearest K Neighbours, Naives Bayes, and Logistic Regression. This allowed my classifier to achieve an accuracy of 93% on the unseen dataset, which was considered excellent.

Ratings

Overall

Overall I enjoyed most of the coursework I completed this year. I found the projects to be quite challenging, but also very rewarding. I ended the year with an average of 90% across all my coursework, which I was pretty pleased with (anything above 70% is a first and considered “Outstanding”). I’m looking forward to next year when I’ll be taking on more challenging projects and learning new things.


Website Source Code