Step 1: Forward Pass

• network makes a guess
• it takes the input, processes it through each layer with its current set of weights and biases, & produces an output

Step 2: Calculate the Error (Loss/Cost Function)

• network compares the output to the correct answer (supervised learning)
• difference between network’s guess and the correct answer is called the ‘loss’ or ‘error’

Step 3: Backward Pass (Back-propagation)

• network uses this error to figure out how much individual weight and bias contributed to the mistake
• it essentially propagates the error backward through the network, from the output layer all the way to the input layer

Step 4: Update Weights

• based on the network’s calculations, the weights and biases are adjusted to reduce the error
• the goal: the next time the network sees this input, it will make a better guess