The Machine Learning “Advent Calendar” Day 18: Neural Network Classifier in Excel

Building a neural network classifier in Excel provides a hands-on understanding of its fundamental mechanisms. The process begins with forward propagation, where the network is defined as a mathematical function that maps inputs to predicted probabilities, demonstrated with a non-linearly separable dataset. This involves structuring a hidden layer with neurons that act as individual logistic regressions. Subsequently, backpropagation is used to train the network by applying gradient descent to minimize the log loss and update the model's weights and biases.