Convert a NumPy array to LaTeX

Sierra de las Nieves (Málaga)

Convert Numpy array to \(LaTeX\)

Sometimes, when you are writing a statistics or a math article, you must go through a tedious process of typing and testing numerous lines of text until your code works. This can be especially boring when dealing with long matrices. If you want to improve your productivity, automating this process comes in very handy, and Python is particularly useful for that.

Functions for matrices and determinants

The functions matrix2latex and det2latex generate an indented \(LaTeX\) string that can be copied or directly rendered in your math article.

The INDENT_SPACES parameter specifies the number of spaces used for indentation.

import numpy as np

INDENT_SPACES = 3

def indent(num_indent=1):
    """
    Number of spaces for indentation
    """
    return num_indent * INDENT_SPACES * " "

def matrix2latex(matrix):
    """
    Convert a NumPy array to LaTeX code as a matrix
    """
    left_latex = r"\left(" + "\n" + indent(1) + r"\begin{array}"
    right_latex = indent(1) + r"\end{array}" + "\n" + r"\right)"
    m_cols = matrix.shape[1]
    array_cols = "{" + "r" * m_cols + "}\n"
    elements_latex = ""
    for row in matrix:
        elements_latex = \
          elements_latex + indent(2) + " & ".join([str(x) for x in row]) + \
            r" \\ " + "\n"
    latex = left_latex + array_cols + elements_latex + right_latex
    return f"$$\n{latex}\n$$"

def det2latex(matrix):
    """
    Convert a NumPy array to LaTeX code as a determinant
    """
    left_latex = r"\begin{vmatrix}" + "\n"
    right_latex = r"\end{vmatrix}"
    m_cols = matrix.shape[1]
    elements_latex = ""
    for row in matrix:
        elements_latex = \
          elements_latex + indent(1) + " & ".join([str(x) for x in row]) + \
            r" \\ " + "\n"
    latex = left_latex + elements_latex + right_latex
    return f"$$\n{latex}\n$$"

Output string

Using the NumPy function np.eye, we can create an identity matrix of the desired dimensions, making it a convenient tool for testing our functions.

print(matrix2latex(np.eye(3, dtype=int)))

$$
\left(
   \begin{array}{rrr}
      1 & 0 & 0 \\ 
      0 & 1 & 0 \\ 
      0 & 0 & 1 \\ 
   \end{array}
\right)
$$

Rendering output in a document

If you are working in a Jupyter notebook, you can render a \(LaTeX\) string using the Math function from the IPython library, which can be imported as follows:

from IPython.display import Math

If you are writing a Python chunk in a Quarto file in RStudio, you need to include the option #| output: asis at the top of the code chunk. This option determines how the output is rendered in the final document.

print(det2latex(np.eye(3, dtype=int)))

\[ \begin{vmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \\ \end{vmatrix} \]

# Identity 4x4 matrix
identity_4 = np.eye(4, dtype=int)
identity_4_latex = r"$$I_4 = " + f"{matrix2latex(identity_4)}"[2:]
print(identity_4_latex)

\[I_4 = \left( \begin{array}{rrrr} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ \end{array} \right) \]

Rendering the output to a file

The \(LaTeX\) string can be rendered and saved to a file using the following function, which is based on the SymPy library.

from sympy import preview

def latex2png(latex, filename, fontsize=300):
    """
    Render LaTeX code to a PNG image
    """
    return preview(latex,
                   viewer='file',
                   filename=filename,
                   euler=False,
                   dvioptions=['-D', f'{str(fontsize)}'])

Other types of automated matrices

Matrices with numbered elements

def element_matrix(n, notation=r"x"):
    """
    Matrix with elements in algebraic notation 
    """
    vec_function = \
      np.vectorize(lambda i, j: notation + "_{" + f"{i + 1}{j + 1}" + "}")
    return np.fromfunction(vec_function,
                           shape=(n, n),
                           dtype=int)

print(matrix2latex(element_matrix(5, notation=r"\theta")))

\[ \left( \begin{array}{rrrrr} \theta_{11} & \theta_{12} & \theta_{13} & \theta_{14} & \theta_{15} \\ \theta_{21} & \theta_{22} & \theta_{23} & \theta_{24} & \theta_{25} \\ \theta_{31} & \theta_{32} & \theta_{33} & \theta_{34} & \theta_{35} \\ \theta_{41} & \theta_{42} & \theta_{43} & \theta_{44} & \theta_{45} \\ \theta_{51} & \theta_{52} & \theta_{53} & \theta_{54} & \theta_{55} \\ \end{array} \right) \]

Triangular matrices

• Upper Triangular Matrix: In an upper triangular matrix, all elements below the main diagonal are zero.

def up_triangular_matrix(n):
    """
    Upper Triangular matrix filled with ones and zeros
    """
    return np.fromfunction(lambda i, j:  1 * np.less_equal(i , j),
                           shape=(n, n),
                           dtype=int)

• Lower Triangular Matrix: In a lower triangular matrix, all elements above the main diagonal are zero.

def lw_triangular_matrix(n):
    """
    Lower Triangular matrix filled with ones and zeros
    """
    return np.fromfunction(lambda i, j:  1 * np.greater_equal(i , j),
                           shape=(n, n),
                           dtype=int)

print(matrix2latex(up_triangular_matrix(5)))

\[ \left( \begin{array}{rrrrr} 1 & 1 & 1 & 1 & 1 \\ 0 & 1 & 1 & 1 & 1 \\ 0 & 0 & 1 & 1 & 1 \\ 0 & 0 & 0 & 1 & 1 \\ 0 & 0 & 0 & 0 & 1 \\ \end{array} \right) \]