EMNIST dataset is MNIST (handwritten digit) dataset with handwritten characters
Applied Persistent Homology and Principal Component Analysis to reduce the dimensionality of dataset. Reduced feature size from 784 (28x28) to 35 while retaining 99% variance
Utilized libraries giotto-ai along-side standard deep learning libraries sklearn, NumPy, Tensorflow/Keras
Achieved 97%-91% training-testing accuracy with a shallow neural network with only 3 hidden layers