Tune Artificial Neural Nets By Hand

Educational tool for exploring neural network tuning both by hand and with backpropagation.

This tool is designed to build intuitions for how ANNs work. The user can tune the weights of a classification ANN manually, by using UI sliders to update weight values, datasets, and activation functions. The results are displayed both as a classification landscape (like the neural network playground) and as a network diagram. The main feature is control over individual weights to help demonstrate the way ANNs compute classifications and why backpropagation is powerful, and what the limits of particular ANN architectures are.