How Perceptrons Work

An artificial neuron takes input values, multiplies each by its corresponding weight, sums the results, and adds a bias.

This calculated value is then passed through an activation function to produce the final output.

Expressed as a formula, it looks like this:

Copy
y = f(w₁x₁ + w₂x₂ + ... + wₙxₙ + b)

• y: Final output of the neuron

• f: Activation function

• w: Weights for each input value

• x: Input values

• b: Bias

Let’s look at a simple scenario: a perceptron deciding whether to “turn on the air conditioner if it’s hot (Input1) and humid (Input2)”.

• Temperature: 86°F (Input1)
• Humidity: 90% (Input2)

• Weight for temperature: 0.7
• Weight for humidity: 0.3

• Bias value: -10

In this case, the perceptron calculates as follows:

(Temperature × Temperature Weight) + (Humidity × Humidity Weight) + Bias
= (86 × 0.7) + (90 × 0.3) + (-10)
= 60.2 + 27 - 10
= 77.2

Since the result does not exceed a set threshold (e.g., 98), the activation function decides to "keep the air conditioner off (Output=0)".

If the result had exceeded the threshold, the decision would be "turn on the air conditioner (Output=1)".

This example shows how a perceptron makes simple decisions by combining inputs, weights, and bias.

When multiple perceptrons are connected, they form a neural network, and when these network layers are stacked to form deep structures, it is known as Deep Learning.

In the next lesson, we will delve deeper into Deep Learning.