Ethan Jordan

Backward Propagation Neural Network

This quarter-long project focused on the design and implementation of a fully functional neural network model using the backpropagation algorithm. Backpropagation is a cornerstone of modern machine learning, enabling neural networks to adjust weights through gradient descent and optimize performance during training.

During this project, I gained hands-on experience in foundational machine learning concepts, including:

• Feedforward and Backpropagation Algorithms: Implemented both the forward pass to compute predictions and the backward pass to minimize error through partial derivatives of the loss function.
• Gradient Descent Optimization: Used mathematical principles like calculus to adjust weights and biases, achieving better accuracy over time.
• Activation Functions: Explored functions such as ReLU, Sigmoid, and Tanh to introduce non-linearity into the model, enhancing its capacity to solve complex problems.
• Overfitting Prevention: Applied techniques such as regularization and dropout to ensure the model generalized effectively to new data.
• Matrix Operations and Vectorization: Optimized performance by implementing efficient matrix multiplications and vectorized computations using Python and libraries such as NumPy.

This project was great exposure to core artificial intelligence concepts, including supervised learning, loss functions, and data normalization. Additionally, I improved my programming skills in Python while using best practices in code modularity, debugging, and unit testing to ensure robustness and scalability.

This experience gave me a solid understanding of how neural networks operate under the hood, which is very useful data science, software development, and security. The problem-solving skills required for this project were next-level and gave me a lot of real life experience in computer science.

Mattercam Project

This project is a real-time dashboard application built using Python and Tkinter that displays live video from Disneyland along with updated ride wait times in a sleek and modern fullscreen interface. The application combines live data scraping, media streaming, and UI design to create a fun and interactive way to check park conditions remotely.

🛠 Key Features & Technologies:

• Tkinter: Designed a custom user interface with themed fonts and layout, styled for a clean, modern appearance using frames and containers.
• Web scraping (BeautifulSoup + Requests): Collected live wait time data for popular Disneyland rides from queue-times.com, parsing HTML to extract relevant ride information.
• Live media streaming (VLC + yt-dlp): Embedded a live YouTube webcam stream from the park using yt-dlp to retrieve the direct video stream and VLC Python bindings to display it inside the GUI.
• Real-time updates: Automatically refreshes wait times every 60 seconds and displays a timestamp to let users know the data is current.
• Dynamic feedback: Rides are color-coded based on their wait time (e.g., green for short, red for long), enhancing user readability.
• Fullscreen mode: Designed for kiosk-style display or immersive viewing.

This project demonstrates:

• Frontend development with Tkinter
• Real-time data automation
• Web scraping and API-less data collection
• Media handling with VLC
• Practical use of external libraries (yt-dlp, BeautifulSoup, vlc)
• Thoughtful UX/UI design principles