How to create an empty matrix with NumPy in Python?

In Python, an empty matrix is a matrix that has no rows and no columns. NumPy, a powerful library for numerical computing, provides various methods to create matrices with specific properties, such as uninitialized values, zeros, NaNs, or ones. Below are different ways to create an empty or predefined matrix using NumPy.

Using numpy.empty()

numpy.empty() function creates an array without initializing its values. This means the contents of the matrix will be arbitrary and depend on the memory state at the time of creation. It is useful when performance is a priority and you plan to overwrite the matrix values later.

import numpy as np

x = np.empty((0, 5))
print("Empty Matrix:", x)
print("Shape of matrix:", x.shape)
print("Data type of matrix:", x.dtype)

Empty Matrix: []
Shape of matrix: (0, 5)
Data type of matrix: float64

Explanation: This code creates an empty matrix with zero rows and five columns. Since there are no rows, the output is an empty list []. The shape of the matrix is (0, 5), which means it has zero rows and five columns. By default, the data type (dtype) of the elements is float64.

Using numpy.zeros()

numpy.zeros() function creates a matrix where all elements are initialized to zero. It is useful when you need a matrix with default zero values, commonly used for initializing weight matrices in machine learning or placeholder arrays in computations.

import numpy as np

z = np.zeros((5, 3))
print(z)

[[0. 0. 0.]
 [0. 0. 0.]
 [0. 0. 0.]
 [0. 0. 0.]
 [0. 0. 0.]]

Explanation: This creates a 5×3 matrix with all elements set to 0. The default data type is float64, but it can be changed using dtype.

Using numpy.full()

numpy.full() function creates a matrix where all elements are initialized to a specified value. This is useful when you need a matrix with a constant value, such as NaN (Not a Number) for missing data representation.

import numpy as np

# nan_matrix
res = np.full((4, 4), np.nan)
print(res)

[[nan nan nan nan]
 [nan nan nan nan]
 [nan nan nan nan]
 [nan nan nan nan]]

Explanation: This creates a 4×4 matrix with all elements as NaN, useful for handling missing values in datasets.

Using numpy.ones()

numpy.ones() function creates a matrix where all elements are initialized to 1. It is useful for situations where you need a default non-zero matrix, such as bias initialization in neural networks.

import numpy as np

# ones_matrix
res = np.ones((3, 3))
print(res)

[[1. 1. 1.]
 [1. 1. 1.]
 [1. 1. 1.]]

Explanation: This creates a 3×3 matrix with all elements as 1.0, stored as float64 by default.