ANNalyze is a Neural Network application built using Ruby that allows users to create and train their own neural network to predict hand-written digits. This project was developed as part of the COS10009 - Introduction to Programming unit at Swinburne University. The application utilizes the Multilayer Perceptron (MLP) model to recognize digits from the MNIST dataset, a popular benchmark for handwritten digit recognition.
The program achieves real-time predictions by training the neural network on the MNIST dataset, providing an educational demonstration of how neural networks work, from training to inference.
This project was awarded a perfect final result of 100%, and I am proud to share it with you as a part of my studies in artificial intelligence and machine learning.
Figure 1: Building Screen
Figure 2: Training Screen
Figure 3: Inference Screen
- Handwritten Digit Recognition: Train and test a neural network to recognize digits from the MNIST dataset.
- Neural Network Training: Build and train a Multilayer Perceptron (MLP) model from scratch.
- Real-Time Predictions: Make predictions in real-time after training the neural network on the MNIST dataset.
- Educational Tool: Learn and visualize how neural networks work through a hands-on project.
Before running the application, make sure you have Ruby installed on your system. You can download it from the official Ruby website.
Additionally, you need to install the following Ruby gems:
gem install numo-narray
gem install ruby2d
gem install savioThese libraries are necessary for matrix operations, graphical display, and saving/loading network models, respectively.
-
Clone or Download the Repository:
git clone https://github.com/yourusername/ANNalyze.git
-
Navigate to the Project Directory: Change to the directory where you downloaded the repository:
cd ANNalyze -
Run the Program: To run the ANNalyze Ruby program, use the following command in your terminal:
ruby main.rb
This project is accompanied by a research paper that explains the Multilayer Perceptron (MLP) model used in ANNalyze for recognizing hand-written digits. The research discusses the key concepts of artificial neural networks, including activation functions, backpropagation, and training on the MNIST dataset.
- 100% Final Result: Achieved a perfect final score of 100% for the Introduction to Programming - COS10009 unit at Swinburne University.
If you found this repository useful or interesting, please feel free to give it a ⭐ (star). Your support is greatly appreciated!
- Swinburne University: For providing an excellent learning experience in programming and artificial intelligence.
- Professors and Teaching Assistants: For their invaluable support and feedback during the course.
- Ruby Community: For maintaining the Ruby programming language and libraries that made this project possible.
- MNIST Dataset: For providing the dataset used to train the neural network for handwritten digit recognition.