Building a Complete Neural Network From Scratch in Microsoft Excel
To keep the spreadsheet manageable, we will build a network to solve a classic binary classification problem: the XOR (Exclusive OR) logic gate. The network will take two inputs, process them through a hidden layer with three neurons, and output a single prediction between 0 and 1. The Architecture 2 nodes ( Hidden Layer ( Z[1]cap Z raised to the open bracket 1 close bracket power A[1]cap A raised to the open bracket 1 close bracket power ): 3 neurons Output Layer ( Z[2]cap Z raised to the open bracket 2 close bracket power A[2]cap A raised to the open bracket 2 close bracket power ): 1 neuron Step 1: Set Up the Training Data
Error = (Predicted Output - Actual Output)^2