Cookie Consent
Hi, this website uses essential cookies to ensure its proper operation and tracking cookies to understand how you interact with it. The latter will be set only after consent.
Read our Privacy Policy


Backpropagation is a popular method used in training artificial neural networks. It involves training a system to refine its predictions or outputs by propagating the error from the output backwards through each layer of the network to modify weights as well as biases. It's a subset of supervised learning algorithms, based on the gradient descent optimization algorithm.

How Backpropagation works

Backpropagation works in two phases: the propagation phase and the weight update phase.

In the propagation phase, the algorithm begins with a forward pass, during which the input data is fed into the neural network to generate predictions. The error between the neural network's prediction and the actual output is calculated using a predefined loss function.

In the weight update phase, the error calculated is then propagated back through the network from the output layer to the input layer. This backward propagation involves the application of the chain rule from calculus to recursively compute the gradient of the loss function with respect to each weight in the network. This process helps to determine the impact or contribution of each weight on the final error.

Subsequently, all weights and biases are updated using this gradient information. The goal here is to adjust the weights and biases so that the total error will be minimized.

This entire process of forward propagation, backward error propagation, and weight and bias adjustment is typically repeated on multiple iterations (or "epochs"), with the goal of reducing the error to an acceptable level.

It's important to note that backpropagation requires the activation functions to be differentiable for it to work properly since the gradients are computed using differentiation.

Lakera LLM Security Playbook
Learn how to protect against the most common LLM vulnerabilities

Download this guide to delve into the most common LLM security risks and ways to mitigate them.

Related terms
untouchable mode.
Get started for free.

Lakera Guard protects your LLM applications from cybersecurity risks with a single line of code. Get started in minutes. Become stronger every day.

Join our Slack Community.

Several people are typing about AI/ML security. 
Come join us and 1000+ others in a chat that’s thoroughly SFW.