All courses
Math · Advanced
Neural Networks
Neural networks power image recognition, translation and chatbots, but underneath they're built from simple maths: weighted sums, a bend, and calculus. Build a neuron that separates data with a line, see why layers are needed, train a model with gradient descent and the chain rule, learn the danger of overfitting, build curves from ReLU kinks, and see how convolutional networks detect edges.
8 lessons~40 min0/8 complete
Start courseHelpful background: Multivariable Calculus
1
Neurons and Networks
From one neuron to many.
2
Learning
How networks find their weights.
3
Inside Modern Networks
Building functions, and how networks see.