NumPy Neural Network

This is a Neural Network written in Python by using the NumPy library.

The model implements backpropagation and Stochastic Gradient Descent (SGD) to learn simple datasets.

Important dependencies are NumPy, Scikit-Learn, Pandas and Matplotlib. The exact versions used in this repo are found in requirements.txt, however, this is an auto-generated file from conda on platform osx-arm64 (Mac M1).

The screenshots below are the results from a neural network with 64 nodes in one hidden layer.

Breast Cancer dataset

Accuracies	Losses

Predictions	Target

5-fold Cross Validation	Confusion Matrix

Iris dataset

Accuracies	Losses

Predictions	Target

5-fold Cross Validation	Confusion Matrix

Moons dataset

Accuracies	Losses

Predictions	Target

5-fold Cross Validation	Confusion Matrix

Wine dataset

Accuracies	Losses

Predictions	Target

5-fold Cross Validation	Confusion Matrix

Digits dataset

Accuracies	Losses

5-fold Cross Validation	Confusion Matrix