Python's **`numpy.fmax()`** computes the element-wise maximum of an array. It compares two arrays and returns a new array containing the maximum values.

> The difference between the `fmax()` and `maximum()` functions in numpy is that the `fmax()` function ignores `NaN` values whereas the `maximum()` function propagates `NaN` values.  

# Syntax
`numpy.fmax()` is declared as follows:

```python
numpy.fmax(x1, x2, /, out=None, *, where=True, casting='same_kind', order='K', dtype=None, subok=True[, signature, extobj]) = <ufunc 'fmax'>
```
In the syntax given above, `x1` and `x2` are non-optional parameters, and the rest are optional parameters. 

> A universal function (ufunc) is a function that operates on ndarrays in an element-by-element fashion. The `fmax()` method is a universal function. 

# Parameters
The `numpy.fmax()` method takes the following compulsory parameters:

* **`x1`** and **`x2`** [*array-like*] - arrays holding the values that need to be compared. If the #key# shape: the shape of an array is the number of elements in each dimension #key# of `x1` and `x2` is different, they must be broadcastable to a common shape for representing the output.   

The `numpy.fmax()` method takes the following optional parameters:


# Return value
`numpy.fmax()` returns the maximum of `x1` and `x2` element-wise. The return type is ndarray or scalar, depending on the input. ﻿﻿Only if both `x1` and `x2` are scalar will the value returned be scalar. 

> If one of the elements being compared is `NaN` (Not a Number), then the *non-nan* element is returned. 

> If both elements being compared are `NaN` (Not a Number), then `NaN` is returned.


# Examples
The examples below show different ways in which `numpy.fmax()` is used in Python. 

The following code example outputs the maximum of two numbers `a` and `b`: 

import numpy as np 

a = 10
b = np.nan
print(np.fmax(a,b))

The example below outputs the result after comparing the arrays `arr1` and `arr2`:

import numpy as np

arr1 = np.array([1.5,3.5,4,np.nan])
arr2 = np.array([np.nan,6.5,2,np.nan])
print(np.fmax(arr1,arr2))

The example below shows the result of comparing the arrays `arr3` and `arr4`:

import numpy as np

arr3 = np.array([[1,2,3], [4,5,6]])
arr4 = np.array([[3,np.nan,1], [6,-5,4]])
print(np.fmax(arr3,arr4))

What is numpy.fmax() in Python?

numpy.fmax() computes element-wise maximum, ignoring NaNs, unlike numpy.maximum() which propagates NaNs.

Parameter	Description
out	represents the location into which the output of the method is stored
where	True value indicates that a universal function should be calculated at this position
casting	controls the type of datacasting that should occur. The same_kind option indicates that safe casting or casting within the same kind should take place.
order	controls the memory layout order of the output function. The option K means reading the elements in the order they occur in memory
dtype	represents the desired data type of the array
subok	decides if subclasses should be made or not. If True, subclasses will be passed through.