This was heavily inspired by Understanding Neural Network Batch Training: A Tutorial, and a surprising amount of the actual code was written by Copilot. The iris.csv file is from the classic Iris Data Set, by way of the article.
The gradient descent and backpropagation code has been preety carefully tested, and any remaining errors are likely errors in my own understanding.
I should consider upgrading this to do some or all of the following:
- Use drop-outs to prevent overfitting.
- Recongnize the MNIST digits.
- Implement ReLU activation functions.
- Implement CNN layers for image processing.
- Try to match the MNIST performance here.
- Implement actual batch learning, just for fun.
Someday.