An grayscale thumbnail arrives as raw pixel brightness values. Two teams build a classifier for it. Both end the same way, with a single output neuron that has its own bias. They disagree about what that neuron should read.
Pipeline A (hand-designed). An engineer writes code that turns the pixels into features chosen by hand: average brightness, number of dark edges, and so on. The output neuron reads those features.
Pipeline B (learned). The pixels go straight into a hidden layer of neurons. Every pixel connects to every hidden neuron, and each hidden neuron has its own bias. The output neuron reads the hidden outputs.
A learned parameter is a number that training adjusts.
How many more learned parameters does Pipeline B have than Pipeline A?
Give the exact integer.