Backed and supported by the National Science Foundation, this project focused on applying the fundamentals of linear algebra and machine learning to the task of handwritten digit recognition. Using Python and various machine learning libraries, we explored different approaches—most notably neural networks—to accurately classify digits from large image datasets. A key part of this experience involved working with UH Mānoa’s new KOA system, a high-performance computing environment that allowed us to efficiently train models and handle large volumes of data.
Throughout the semester, I gained hands-on experience with essential machine learning tools and techniques, including model training, data preprocessing, and performance evaluation. Using the KOA system gave me valuable exposure to computing resources typically used in real-world AI applications, and allowed for faster experimentation and iterative improvements. I strengthened my programming skills in Python, deepened my understanding of neural networks and linear algebra, and learned how to interpret and refine model outputs. Additionally, the collaborative nature of the project helped build my teamwork and problem-solving skills in a fast-paced, technically challenging environment.