Building and training a Neural Network from scratch in Substance Designer

Practice/play 😊 ❤️ #SubstanceDesigner
Substance Designer learns its first math function from data points via a Neuronal Network implemented from scratch. This simple model uses 12 ReLU functions to fit nonlinear relationships between data points. Gradient descent converges fast as this is a simple convex optimization problem. Gradient descent finds the global minimum in a space with 12 dimensions and learns 12 parameters.