Inputs, Weights, Bias, and Neurons
Learn how small pieces of information become one decision.
Inputs, Weights, Bias, and Neurons explained visually
Use an everyday umbrella decision to look inside one artificial neuron.
Inputs provide the clues
An artificial neuron receives numerical pieces of information called inputs. In an umbrella decision, those inputs might represent sky darkness, rain probability, and wind speed. A real network may receive thousands or millions of inputs, but the principle is the same: each input describes one part of the example.
Weights control importance
A weight tells the neuron how strongly an input should influence the result. A reliable rain forecast may receive more weight than wind speed. The neuron multiplies each input by its weight, then adds the weighted values to create one combined score.
Bias moves the decision point
Bias is an additional learned value. It lets a neuron shift its decision even when the inputs are small. Someone who strongly dislikes getting wet may take an umbrella with weaker evidence; the bias gives an artificial neuron similar flexibility.
A neuron sends one useful signal
After combining inputs, weights, and bias, the neuron applies an activation function and sends an output forward. Training adjusts the weights and bias from examples rather than requiring a person to choose every value manually.
Quick check
What is an input?
A numerical piece of information provided to a neuron.
What does a weight control?
How strongly an input influences the neuron's result.
Why is bias useful?
It shifts the decision point and gives the neuron more flexibility.
Lesson 2 transcript
Imagine that you are deciding whether to take an umbrella when you leave home. You check several clues. Is the sky dark? Is rain shown in the forecast? Is the wind getting stronger?
An artificial neuron makes a decision in a similar way. It receives pieces of information called inputs. In our umbrella example, one input can represent the darkness of the sky. Another can represent the chance of rain. A third can represent the wind speed.
The neuron does not treat every input equally. Each input has a weight. A weight tells the neuron how important that input should be. The rain forecast may receive a strong weight because it gives useful information. Wind speed may receive a smaller weight because wind does not always mean rain.
The neuron multiplies each input by its weight and adds the results. This creates one combined value. You can think of it as a score that summarizes all the clues.
The neuron may also add a value called bias. Bias shifts the decision point. Suppose you strongly dislike getting wet. Even when the evidence for rain is uncertain, the bias can make you more likely to take the umbrella. In a neural network, bias gives the neuron flexibility. It helps the neuron make useful decisions even when every input is small or zero.
After combining the weighted inputs and bias, the neuron passes the result through an activation function. This determines what signal the neuron sends forward.
During training, the network learns better values for its weights and biases. We do not normally choose all these values by hand. The network adjusts them as it studies examples and receives feedback about its mistakes.
The main idea is simple. Inputs provide information. Weights control importance. Bias shifts the decision point. The neuron combines them and produces an output.
In the next lesson, we will examine the activation function, the gate that shapes the neuron's final signal.