# Learning Rate

> How big each training step is. Too big jumps around, too small crawls.

**Field:** Artificial Intelligence  
**Category:** Machine Learning  
**Also known as:** Step Size

## Think of it like

Think of it like the size of your steps down a steep hill. Too big and you trip.

## Definition

The learning rate is a setting that decides how big each training step is. With a large learning rate, the weights change a lot at each step. With a small one, they change only a little. Too large, and training jumps around and never settles. Too small, and training takes far too long. Common values are small, like 0.001.

## Example

Emma trains with a learning rate of 0.5, and the loss bounces wildly. She drops it to 0.001, and the loss falls smoothly.

## Why it matters

It is often the single most important setting to tune. The wrong value can waste days of training.

## Common mistake

Keeping it the same from start to finish. Many teams start larger and shrink it as training goes on.

## Three principles

- **Step size.** How far each update moves.
- **Too big.** Jumps and never settles.
- **Too small.** Slow, and may stall.

## Related terms

- [Gradient Descent](https://learn.benibix.com/term/ai/gradient-descent)
- [Hyperparameter](https://learn.benibix.com/term/ai/hyperparameter)
- [Epoch](https://learn.benibix.com/term/ai/epoch)
- [Loss Function](https://learn.benibix.com/term/ai/loss-function)
- [Optimizer (Machine Learning)](https://learn.benibix.com/term/ai/optimizer)

## Taught in

- [Gradient Descent and Backprop](https://learn.benibix.com/learn/ai-machine-learning/gradient-descent-and-backprop): Machine Learning, 14/18

![A U-shaped curve of loss. A ball takes small steps down the slope to the bottom. On a second curve, steps that are too big jump from side to side past the bottom.](https://learn.benibix.com/diagrams/nn-gradient-descent.svg)

---

Source: https://learn.benibix.com/term/ai/learning-rate
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-17
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
