Exporing machine learning.
This code implements a simple feedforward neural network in Python, using components like activation functions, loss functions, weight and bias initializers, and optimizer algorithms. The network provides a foundation for training and evaluating models using backpropagation and optimization techniques. It is meant to be a starting place for making simple neural networks, and is easily expandable.
A simple test for ml.py on the mnist data set
- It takes ~9 seconds to train the model
- Achieves ~92.5% testing accuracy
- I've gotton the testing accuracy to 97%, but it takes longer to train