I try to learn something about machine learning! I created a neural network with 3 layers to recognize hand drawn digits. It's trained on MNIST data from Kaggle: https://www.kaggle.com/code/hojjatk/read-mnist-dataset. For the activation function I used ReLU and for the output activation function I used Softmax. For loss I used cross entropy to calculate it and for the optimizer I used SGD. All of this is written in plane python except for numpy for the math..