Weights (in neural networks)
Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary
The numeric parameters of a neural network that set how strongly each input influences each neuron, learned during training to reduce prediction error.
What are Weights?
Weights are the numbers a neural network learns during training: each one sets how strongly one neuron’s output influences the next neuron. Everything a trained network “knows” is stored in its weights.
That makes them the network’s memory in the most literal sense. Copy the weights and you have copied the model; this is why the “parameters” quoted for large models (7 billion, 70 billion) are simply weight counts.
How Weights Work
A neuron computes a weighted sum of its inputs, adds a bias term, and applies an activation function:
\[ y = f(w_1 x_1 + w_2 x_2 + \dots + w_n x_n + b) \]A large positive weight means that input pushes the neuron hard toward firing; a weight near zero means the input is ignored; a negative weight suppresses it.
Training is the search for good weight values. The network starts from random weight initialization, makes predictions, and measures its error with a loss function. Backpropagation then computes each weight’s share of the blame, and gradient descent nudges every weight against its gradient:
\[ w \leftarrow w - \eta \frac{\partial L}{\partial w} \]where \( \eta \) is the learning rate. Millions of small nudges later, the weights encode the patterns in the training data.
Example of Weights
Consider a small network that flags emails as spam, with one input per word in its vocabulary. At initialization, the weight linking the word “lottery” to the spam output is a random value near zero, so the word carries no signal.
A spam email containing “lottery” arrives and the network wrongly clears it. Backpropagation traces the error to the weights involved and increases the “lottery” weight slightly.
After thousands of such corrections, that weight has grown large and positive, while weights for neutral words like “meeting” have settled near zero. Nobody told the network which words matter; the weights absorbed it from the mistakes.
Related AI terms: Backpropagation · Gradient Descent · Neural Network · Weight Initialization · Weight Decay
Did you like the Weights (in neural networks) gist?
Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.
Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About