I've been working on my customizable neural network project as a way to learn more about machine learning and effective data visualization. Initially, the program simply created and trained a neural network using the Keras library on a synthetic dataset for binary classification and included an evaluation of the model on a test dataset and a visualization of the decision boundary of the trained model.

I've officially released the updated version that accomplishes my original goal of having this program be an educational tool! It now prompts the user to customize the neural network model (using different optimizers, callbacks, and different values for numbers of hidden layers, learning rate, etc.), explains the meaning and use case for each feature, and finally implements the choices, displaying how they affect the neural network's ability to learn. It lists all customizations chosen in the final visualizations so you can keep track of how your choices affect the model's ability to learn.

The current repository includes the updated project code (haywire.py), the requirements.txt file to download all the dependencies, as well as a video demonstration (the .mkv file) and samples of the plots produced. The haywire_example.py file is the original project, before I made it customizable.