When you check your prediction column, you will see that the outputs closely match the target XOR logic: rows with identical inputs will output near 0 , and rows with mixed inputs will output near 1 . You have successfully built, visualized, and trained a neural network entirely inside a spreadsheet. Share public link
To keep the spreadsheet clean and understandable, we will build a network with the following architecture: Accepts two numerical features (
This is where the magic happens. We will use MMULT (matrix multiplication) for efficiency. Step 1: Calculate Hidden Layer Nodes For each row, calculate the weighted sum plus bias: build neural network with ms excel new
Sigmoid(x)=11+e−xSigmoid open paren x close paren equals the fraction with numerator 1 and denominator 1 plus e raised to the negative x power end-fraction In Excel: =1 / (1 + EXP(-Final_Sum)) . 5. Training the Network (The "New" Excel Part)
Set to your Weight and Bias blocks ( Weights_1, Bias_1, Weights_2, Bias_2 ). Select GRG Nonlinear as the solving method. When you check your prediction column, you will
In cell K6 , put the Target (Y). For XOR, the target is 1 if inputs are different, else 0 . In cell L6 (Loss): =(I6# - K6)^2 (Instantaneous error) In cell L8 (Total MSE): =AVERAGE(L6#)
Building a neural network in MS Excel is a feasible task, although it may not be the most efficient or scalable approach. By using Excel's built-in functions and tools, you can create a simple neural network that can learn from data. However, for more complex neural networks or larger datasets, you may want to consider using specialized machine learning software or libraries. We will use MMULT (matrix multiplication) for efficiency
By toggling your Reset_Cell from TRUE to FALSE , holding down the key will force Excel to recalculate, updating the weights and training the network live before your eyes. 7. Why Build a Neural Network in Excel?
Loss=12(A2−Y)2cap L o s s equals one-half open paren cap A sub 2 minus cap Y close paren squared
: Instead of copying formulas down thousands of rows, a single formula can now "spill" an entire layer of calculations across the grid, making the architecture of a Multi-Layer Perceptron (MLP) much easier to manage. Python in Excel
allow you to encapsulate the complex math of a neuron—weights, biases, and activation functions—into a single, reusable formula. Dynamic Arrays