Minor: Computer Science and Film
Faculty Mentor: Dr. Lynette Boos, Mathematics and Computer Science
We designed two neural networks that can learn how to classify three different types of partial differential equations (PDEs). Our data consists of numerical solutions to three categories of PDEs: Burger’s, Diffusion, and Transport equations. Using TensorFlow and the Keras library, we performed two tasks – the first a binary classification of Burger’s and Diffusion equation data, and the second a multi-label classification incorporating the Transport Equations as well. Our binary classification network requires vector labels to perform efficiently. Furthermore, our tertiary classification network continues to show that vector labeling provides the most accurate predictions. Our networks consistently make more accurate classifications and predictions than other classification tools, particularly Classification and Regression Trees (CART) and Support Vector Machine (SVM).
4-22-2020 12:00 AM