Demo · a real Tessorio lesson
Here is a real one. Press play.
Read this lesson as text
How Neural Networks Work
How Neural Networks Work
Let's look at how a neural network actually works, starting from a single neuron and building up to a full network.
Three Layers, One Pipeline
A network is built in layers. Data enters through the input layer, gets transformed by the hidden layers in the middle, and comes out the other side as a prediction at the output layer.
Check-in: Where does a network's prediction come out?
Where does a network's prediction come out?
Perceptron vs. Neural Network
Compare that with the original perceptron from 1958. It had just one layer, so it could only separate things with a straight line. Networks with two or more layers, trained with backpropagation since 1986, can learn curved, complex boundaries.
Check-in: Why can a neural network learn curved boundaries when a perceptron can't?
Why can a neural network learn curved boundaries when a perceptron can't?
One Forward Pass
Every neuron repeats the same three moves. First, it multiplies each input by a weight and adds them all up. Second, it adds a bias, which shifts that sum up or down. Third, an activation function, like ReLU, decides how much of that signal passes forward to the next layer.
Check-in: Which part decides how much signal passes forward?
Which part decides how much signal passes forward?
What to watch for
Three things to notice.
The picture draws itself
Each part of the diagram appears on the exact word that names it, so what you see and hear line up.
It stops and asks
The lesson waits for your answer before it moves on, and tells you straight away whether you got it.
Every word on screen
The words light up under the picture as they're spoken, so you can read along or catch what you missed.
Want lessons like this for what you're learning?
Start with a free account, and pick a plan when you're ready.