Neural Networks by Hand in Excel
Neurons, activation functions, a full forward pass, loss, backpropagation, learning rates, softmax and early stopping, computed cell by cell so you see what PyTorch and TensorFlow do inside.
The first lesson is free — no account needed.
About this course
Build a neural network in Excel and see what PyTorch and TensorFlow do inside, because every number sits in a cell you computed. Using one coffee club's renewal data, you build a single neuron with a weighted sum and bias, compare step, sigmoid, tanh and ReLU activations, run a forward pass through a 2-2-1 network, score predictions with MSE and cross-entropy, do one full backpropagation step with every gradient from the chain rule, watch three learning rates crawl, converge or explode, turn raw scores into softmax probabilities and pick the best epoch with early stopping. No coding required, and the first lab is free.
Lessons
- A Single Neuron: Weighted Sum + BiasFree
- Activation Functions: Step, Sigmoid, tanh & ReLU Pro
- Forward Pass of a 2-2-1 Network Pro
- Loss Functions: MSE vs Cross-Entropy Pro
- Backpropagation: One Training Step by Hand Pro
- Learning Rate: Five Steps, Three Speeds Pro
- Softmax & Multi-Class Probabilities Pro
- Overfitting: Train vs Test Error & Early Stopping Pro