Matlab is probably more comparable to R than to Python, in the sense that it is a tool created by and for specialists in scientific and statistical computing - i.e. people with advanced degrees in Engineering and the Sciences, who make up perhaps one percent or less of the word's population (actually, probably only a few percent of the population at an average university). Python is a popular language because it is simple, general, an widely used to teach introductory programming, so everybody knows it. Because there is such a large installed based of users, it is natural that there are many users of numpy, scipy, and sklearn. Matlab is also very expensive, but it makes up for that by having far better documentation than Python and code for signal processing and High-Performance Computing that far better written and tested. For example, the parallel toolbox and signal processing toolbox (as well as neural networks toolbox) when used together will fully exploit all the available cores and/or GPUs of your system transparently. This is something that can take weeks or months to get working on python (ask someone who has tried to do multi-GPU Tensorflow using Python on Windows, or even on Linux). Python in fact has some limitations on concurrency baked in, but I am not up to speed on what they are.

For the biomedical sciences, R is probably the alternative to Matlab, as it is likewise written by and for specialists. Like Matlab it was intended from the start for high-performance vectorized computing, and like Matlab one of its defining features is that in part a wrapper for the core LAPACK libraries, which are written in FORTRAN (...which explains why both Matlab and R store arrays in column major). R has also benefited from a tradition of strict documentation requirements, which include the stipulation that libraries contain running demo code and if necessary, data sets (there are many sample data sets build into the R language). Many of the routines in sklearn, for example, as well as the general structure of PANDAs, which are the core data science tools in Python, are essentially ports and/or modifications of what was originally R code. There is a general large-scale effort among the bio-statistics community to publish, preserve, and - lately - make reproducible the knowledge of their field in the form of R packages. If you are doing research in that field, it means you can not only use the techniques that so-and-so published, you can use their code and test your modifications to it on their original data. The downside of R is that it is built to make things that are super-hard and complex easy, at the cost of having syntax and programming paradigms that are not intuitive to the newcomer, but payoff after investing in the steep learning curve.

So, net-net, Matlab and R are more powerful languages for the fields you mention, but Matlab is expensive and R is hard to learn. Python is easy to learn, but lacks uniformity in code, quality, and documentation. However, let's get Bayesian: the fact that Python is far more popular a priori will lead you to experience the phenomenon of hearing many people recommend it for ... anything. Some will point out that "it is the most popular language for Data Science" without conditioning that statistic on the population of users, in other words, putting the postdoc in biostatistics in the same bin with all those people who take a "free data science bootcamp." Chances are, the person making that kind of unqualified recommendation isn't an expert in the field.

Also, I'd add to #Jamie Matthew Bright's answer above, that you have to be super-careful getting answers from Stack Overflow (including for R and Matlab) - as many of the answers, despite being upvoted can be wrong or bad practice. Just because something is popular doesn't mean it's good.

W

I recently learned Python for spatial data analysis, so my answer is generic to the difference, not specific to signal processing. The key difference I find between the languages is the extent of the documentation and therefore speed of learning of the Matlab functions when compared to the Python libraries that may or may not have documentation.

There are, however, significantly more contributors to the Python help on places like StackOverflow.

Both are great programming languages. If you are good in 1, it is generally easier to learn the other. The key difference for me is, Python is opensource (free). Whereas MATLAB requires a license fee. Both languages have (more or less) equal amount of online support and active forums.

Matlab is probably more comparable to R than to Python, in the sense that it is a tool created by and for specialists in scientific and statistical computing - i.e. people with advanced degrees in Engineering and the Sciences, who make up perhaps one percent or less of the word's population (actually, probably only a few percent of the population at an average university). Python is a popular language because it is simple, general, an widely used to teach introductory programming, so everybody knows it. Because there is such a large installed based of users, it is natural that there are many users of numpy, scipy, and sklearn. Matlab is also very expensive, but it makes up for that by having far better documentation than Python and code for signal processing and High-Performance Computing that far better written and tested. For example, the parallel toolbox and signal processing toolbox (as well as neural networks toolbox) when used together will fully exploit all the available cores and/or GPUs of your system transparently. This is something that can take weeks or months to get working on python (ask someone who has tried to do multi-GPU Tensorflow using Python on Windows, or even on Linux). Python in fact has some limitations on concurrency baked in, but I am not up to speed on what they are.

For the biomedical sciences, R is probably the alternative to Matlab, as it is likewise written by and for specialists. Like Matlab it was intended from the start for high-performance vectorized computing, and like Matlab one of its defining features is that in part a wrapper for the core LAPACK libraries, which are written in FORTRAN (...which explains why both Matlab and R store arrays in column major). R has also benefited from a tradition of strict documentation requirements, which include the stipulation that libraries contain running demo code and if necessary, data sets (there are many sample data sets build into the R language). Many of the routines in sklearn, for example, as well as the general structure of PANDAs, which are the core data science tools in Python, are essentially ports and/or modifications of what was originally R code. There is a general large-scale effort among the bio-statistics community to publish, preserve, and - lately - make reproducible the knowledge of their field in the form of R packages. If you are doing research in that field, it means you can not only use the techniques that so-and-so published, you can use their code and test your modifications to it on their original data. The downside of R is that it is built to make things that are super-hard and complex easy, at the cost of having syntax and programming paradigms that are not intuitive to the newcomer, but payoff after investing in the steep learning curve.

So, net-net, Matlab and R are more powerful languages for the fields you mention, but Matlab is expensive and R is hard to learn. Python is easy to learn, but lacks uniformity in code, quality, and documentation. However, let's get Bayesian: the fact that Python is far more popular a priori will lead you to experience the phenomenon of hearing many people recommend it for ... anything. Some will point out that "it is the most popular language for Data Science" without conditioning that statistic on the population of users, in other words, putting the postdoc in biostatistics in the same bin with all those people who take a "free data science bootcamp." Chances are, the person making that kind of unqualified recommendation isn't an expert in the field.

Also, I'd add to #Jamie Matthew Bright's answer above, that you have to be super-careful getting answers from Stack Overflow (including for R and Matlab) - as many of the answers, despite being upvoted can be wrong or bad practice. Just because something is popular doesn't mean it's good.

W

Signal processing toolbox in MATLAB helps in signal processing. But in python environment it is easier to develop any functions/procedures based upon the need.

That is,in MATLAB, many built-ins available and are also easy to use .

And , in PYTHON, any customization is easier.Another adv is its an OPENSOURCE

Keep in mind :

julia (https://julialang.org/ ) is an open source programming language that is faster than python & by far from MATLAB.

New packages are being added, in addition to having the capability to call functions from python & R.

In general, I recommend switching & contributing to open source languages.

Their progress is fast & unparalleled.

As Microsoft wants to advertise against Linux, MATLAB also wants to do the same on open source softwares.

I think MATLAB is not well-suited for scientific programming for the following main reasons:

• There is no systematic handling of modules & dependencies. For example, if you run someone's matlab code it may happen that some function is not on your 'matlab path'. One may not even be aware of this because the modules are not imported at the beginning of a script (like library(RcppCNPy) in R or import numpy as np in python). This is a serious problem for collaboration & reproducibility (if person A and person B have a similar function with the same name on the 'matlab path', the difference may never be noticed...).
• There is no central package repository and systematic handling of versions. E.g. you can't specifically pin a dependency to version 1.2.3 of example_package and fetch it easily (like pip install example_package==1.2.3). So even if persons A and B have the same example_package, there is no mechanism to make sure they have the same version (expect manually sending around stuff which is error-prone). Very bad for reproducibility.
• It's proprietary and a new version is released twice a year (for more money). So it's difficult to reproduce results (e.g. person A has version 2015a, person B has version 2017b). New matlab version resulted in visibly different fMRI preprocessing results halfway through an analysis. Very very bad for reproducible and consistent analysis.
• Besides that, outside of academia few people are willing to pay the ridiculous licence fee when there are better alternatives, so perhaps not an ideal choice for anyone considering a career outside of academia. Finally, alternatives (especially python and R) have much larger communities with more dynamics. I suspect that machine learning will finally finish off matlab. mathworks simply can't compete against google's tensorflow which is free & open source and technically already by far superior.

No offence meant, I absolutely understand why it makes perfect sense to continue using existing matlab code (safely baked into a docker image to make sure the results don't change with every version update).

I just don't think it's a good idea to ride the dead horse any further for new projects.

I work with EEG signals and most of the toolboxes I need are in MATLAB: EEGLAB, ERPLAB, FieldTrip, Brainstorm, etc., all run in MATLAB.

Phyton is free. In my opinion, matlab since is righnow a common software in the academic and industry area is better

In my opinion, when it comes to signal processing, MATLAB is the most useful one right now (there are more toolboxes available in MATLAB), but Python will be the future. Maybe it's similar to how R has overtaken SPSS when it comes to statistical analysis during the past 15 years.