• The project has been worked on for about 6 months with the help of a lecturer, Mr Gi Soong Chee and two JC students, Rui Yang and Isaac
• The game allows for the user to pick a class (out of about 5 others) and lets them travel around a big map, defeating monsters and getting loot to become stronger
• I built the game and published it to the BuildingBlocs Discord server and maintained it for 2 months before shutting it down due to exams
• I talked about the project during PyconSG 2021
• I learnt about asyncio and got used to asynchronosity through the project
• I also used my own written module, FileMonster, to read and write all game files
• I had made codes for parts of machine learning to be used together to make a neural network from scratch
• The codes include normalization, KNN, mean/std calculation and convolutional/pooling layers all built from scratch using Numpy
• I had also made code for gradient descent
• This was my first game and first web development project so it was my first experience with HTML, CSS and JS
• I learnt along the way while developing this game and it took about 2 weeks to finish
• In this game you had to click a button to "mine" and you would get a random amount of random ores that have different rarities. Then you could sell them for a fixed amount of money which you can use to purchase upgrades such as increased amount of ores mined each time or decreased rarity in ores.
• There were features such as automining as well which costs an "automine card" which you can get randomly from mining with a low chance of dropping, this makes use of asynchronosity
• This was my first deep learning project and I'd decided to do a simple CNN (Convolutional Neural Network) using the tensorflow library
• I used a basic CNN architecture and fed it pre-processed images of 4 different groups of blood cells
• The CNN could then predict the type of bloodcell from an image
• I built a perceptron from scratch using only numpy
• I tested my perceptron with already established biases and weights of NOT, AND and XOR perceptrons
• I also trained the perceptron with inputs and all accuracy was 100%