How a network computes, why gradients vanish, and what makes depth trainable
One unit draws one straight fence. That is not a limitation of the training rule, or of how long you train, or of the learning rate. It is what the formula w · x + b is. This lesson finds a problem with four data points that no straight fence can solve, and proves that no weights exist for it.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ What One Neuron Cannot Do