Easily integrate Custom Functions in MATLAB with Python (codes included)

Introduction

MATLAB implementation are usually quite reliable as it is developed by the professionals. But the advantages of using Python is immense. In this post, I will show how you can benefit the MATLAB function in your Python script.

Defining custom function in MATLAB

Let us make a custom function in MATLAB that we can use in Python. For the purpose of demonstration, I will use a very simple function but the same idea applies to any function.

Eigenvalues and eigenvectors in MATLAB

function [V,D] = eigFunc(A)
%returns diagonal matrix D of eigenvalues and matrix V 
% whose columns are the corresponding right eigenvectors, 
% so that A*V = V*D.
[V, D] = eig(A);
end

I saved the above function as eigFunc. This function takes in a square matrix as input and outputs diagonal matrix D of eigenvalues and matrix V whose columns are the corresponding right eigenvectors [see eig function in MATLAB].

Let’s first use this function in MATLAB for test purpose.

clear; close all; clc
A = gallery('lehmer',4);

[V,D] = eigFunc(A)

This returns:

V =

    0.0693   -0.4422   -0.8105    0.3778
   -0.3618    0.7420   -0.1877    0.5322
    0.7694    0.0486    0.3010    0.5614
   -0.5219   -0.5014    0.4662    0.5088


D =

    0.2078         0         0         0
         0    0.4078         0         0
         0         0    0.8482         0
         0         0         0    2.5362

This works great in MATLAB as expected (coz the function is exact copy of the eig function in MATLAB. But how can we use this function in Python?

Call MATLAB in Python

There are two routes, and which one you pick depends on whether the person running the code has MATLAB installed:

Using MATLAB Engine API for Python

The easiest way to use the matlab function in Python is by using the matlab.engine. You can install matlab library by following these two ways.

Install from inside MATLAB:

cd (fullfile(matlabroot,'extern','engines','python'))
system('python setup.py install')

Install directly from terminal:

Navigate to the MATLAB source location and compile the python. For installing matlab library in Mac:

cd /Applications/MATLAB_R2021b.app/extern/engines/python/
python setup.py install

Please note that this MATLAB engine API will be installed for the the specific version of python. If you are using anaconda, you can inspect the version of Python you are installing the MATLAB engine API for. It is essentially the same way you install other Python libraries. For details, visit here.

Simpler install since R2022b: use pip. You no longer have to run python setup.py install from inside matlabroot (and setup.py install is itself deprecated by modern setuptools). The engine is now on PyPI, so you can just run pip install matlabengine. The catch is version matching: the matlabengine release must match your installed MATLAB release — e.g. pip install matlabengine==25.1.* for R2025a — and each release supports a specific band of Python versions (the current package supports Python 3.9–3.13). Pip will list the compatible versions if you ask for one that does not match. The setup.py method above still works for older MATLAB releases.

import matlab
import matlab.engine
eng = matlab.engine.start_matlab()
A = [[1.0000,    0.5000,    0.3333,    0.2500],
     [0.5000,    1.0000,    0.6667,    0.5000],
     [0.3333,    0.6667,    1.0000,    0.7500],
     [0.2500,    0.5000,    0.7500,    1.0000]]
A = matlab.double(A)
V, D = eng.eigFunc(A, nargout=2)
print("V: ", V)
print("D: ", D)

eng.quit()

Here is the output:

V:  [[0.06939950784450351,-0.4421928183150595,-0.8104910184495989,0.37782737957175255],				[-0.3619020163563876,0.7419860358173743,-0.18770341448628555,0.5322133795757004],[0.7693553355549393,0.04873539080548356,0.30097912769034274,0.5613633351323756],[-0.5218266058004974,-0.5015447096377744,0.4661365700065611,0.5087893432606572]]

D:  [[0.20775336892808516,0.0,0.0,0.0],[0.0,0.40783672775946245,0.0,0.0],[0.0,0.0,0.8482416513967358,0.0],[0.0,0.0,0.0,2.536168251915717]]

The results are same as before. The nargout argument tells the matlab based function to output 2 results here.

Create Python Package: Using MATLAB Compiler & MATLAB Runtime

This is all well and good if you have MATLAB installed in your system. But what if you wanna give your Python script so someone who does not have MATLAB installed on their system. In that case, you can build a Python library using the library compiler app in MATLAB. For details, visit Generate a Python Package and Build a Python Application.

MATLAB Runtime installation

Please note that, the user need to install MATLAB runtime for successfully using this library. MATLAB runtime helps in running compiled MATLAB applications or components without installing MATLAB. The runtime can be downloaded from here for Windows, Mac and Linux OS and it is free to download.

Import MATLAB based library for Python

import eigFunc
eigFuncAnalyzer = eigFunc.initialize() #calls the matlab runtime
A = [[1.0000,    0.5000,    0.3333,    0.2500],
     [0.5000,    1.0000,    0.6667,    0.5000],
     [0.3333,    0.6667,    1.0000,    0.7500],
     [0.2500,    0.5000,    0.7500,    1.0000]]
A = array.array(A) %not tested
V, D = eigFuncAnalyzer.eigFunc(A, nargout=2)
print("V: ", V)
print("D: ", D)
eigFuncAnalyzer.terminate()

Please note that you can design your eigFunc.m in a way that you can simply load the mat data and you can use scipy.io.savemat function to save the python data to mat format. For details see the scipy documentation.

Data Type Conversions

Recap

• The MATLAB Engine API for Python (import matlab.engine, eng = matlab.engine.start_matlab()) drives a live MATLAB session from Python — quickest to set up, but every user needs MATLAB installed and licensed.
• Since R2022b the engine installs with a plain pip install matlabengine, version-matched to your MATLAB release (no more setup.py install).
• Pass data across the boundary with MATLAB’s typed wrappers such as matlab.double(...), and use nargout to tell a function how many outputs to return.
• To deploy to users without MATLAB, package the function with MATLAB Compiler SDK and have them install the free MATLAB Runtime — no license required.
• Refer to the data-type conversion table above (double↔float, struct↔dict, cell↔list/tuple, …) whenever results look wrapped in an unexpected type.

Where to go next

• Time-frequency analysis in MATLAB — a MATLAB workflow you could now drive from Python.
• Test your scientific hypothesis in MATLAB — another MATLAB routine worth bridging.
• Install the MATLAB Engine for Python — the official, version-by-version compatibility reference.

References

1. How to Call MATLAB from Python
2. Call MATLAB Functions from Python
3. Generate a Python Package and Build a Python Application