**NumPy**

*chevron_left*

**Cookbooks**

# Flattening NumPy arrays

*schedule*Aug 11, 2023

*toc*Table of Contents

*expand_more*

**mathematics behind data science**with 100+ top-tier guides

Start your free 7-days trial now!

We can reduce a n-dimensional NumPy array to 1D using either the `flatten()`

or `ravel()`

method available to all NumPy arrays. We first demonstrate their usages, and subsequently their differences.

# Example

Consider the following 2 by 3 array:

```
array([[1, 2, 3], [4, 5, 6]])
```

We can use the `flatten`

method as follows:

```
array([1, 2, 3, 4, 5, 6])
```

We can also use the `ravel`

method:

```
array([1, 2, 3, 4, 5, 6])
```

Notice the outputs are the same - we end up with a one-dimensional flattened array.

# Difference between flatten and ravel

The `flatten(~)`

returns a separate copy of the NumPy array. This means that making modification on the original array would not have any impact on the flattened array. Just to illustrate, study the following code:

```
```

On the other hand, the `ravel(~)`

returns a NumPy array that shares the same memory address as the original array:

```
```

As arrays `x`

and `y`

both share the same memory address, when we update value `[0,0]`

in array `x`

, we can see that the new assignment of `9`

is reflected in array `y`

also.

In terms of speed and memory-savings, `ravel()`

is superior. Therefore, use `ravel()`

if you are certain that:

you won't need the original array

you won't make any modification on the original array

You should also be reminded that, unless we are dealing with large amounts of data, the performance difference is negligible. Therefore, you may want to use `flatten()`

to ensure that nothing out of the ordinary happens.