This project achieved 96% training data accuracy by implementing a self-developed multi-layer perceptron (MLP) neural network with feed-forward and back-propagation algorithms, without utilizing machine learning libraries like PyTorch or TensorFlow.
git clone https://github.com/wallinslax/handwritten-digits-classifier.git
cd data
python mnist_csv3.py
cd ..
python mlpForHandwrritenDigit.py