Well.. calling Python as an open source alternative to Matlab is ... I think not correct..

Octave is such alternative.. Python in general has much broader scope and it is very much widely used....(see Google)..

Well.. calling Python as an open source alternative to Matlab is ... I think not correct..

Octave is such alternative.. Python in general has much broader scope and it is very much widely used....(see Google)..

Compare Python with R: the easy of R packages for all OS platforms cannot compare with the relevant of Python packages. Besides that do not forget that Matlab as a commercial product (ie Mathworks) spends a lot of money to promote it at Universities. It is a big story why Matlab has gained so much attention, although it is slower than Fortran.

I see no reason why, Python and R are equally good as Matlab... switch to these alternatives, and publish good papers that cite them... :)

Python is one of the simplest languages to learn and having lot of functionality with it,

As you observe Matlab is mainly used in simulations, optimizations etc, So I would say Python is still more powerful that Matlab, and even a lot cheaper

I would say that it is being used! In my last place of work, we used Matlab for doing some initial analysis, purely because it's easier to look at data and produce some visualisations quickly (for me anyway), but it simply wasn't an option for a production system, partly due to licensing but also as we had to integrate it with existing web services (and provide a restful interface). Python was the language of choice for the production system (together with Django), and all of the functionality that was present in the Matlab version was present in the Python version. I think that in academic circles, people tend to use Matlab as they tend to have site-wide licenses (or at least departmental licenses), and why wouldn't they? It's a slick product. However, in the "real" world, choices are often made for licensing or other architectural reasons, and many companies will often opt to use the technology that is already most prevalent in their products (i.e. if you are a Java house, you're more likely to choose Java), as this causes the least pain for those in charge of deployment. Personally this is where I see python as having the greatest trouble, as the deployment of python packages is still immature when compared with other languages.

Being a student, cost shouldn't be a barrier to me, but I am a huge supporter of the open source movement and love the fact that its all free. Besides coding in Python feels like a breeze, since it looks good. As famous entrepreneur and investor, Paul Graham puts it: "That may seem a frivolous reason to choose one language over another. But it is not so frivolous as it sounds: when you program, you spend more time reading code than writing it. You push blobs of source code around the way a sculptor does blobs of clay. So a language that makes source code ugly is maddening to an exacting programmer, as clay full of lumps would be to a sculptor." and for some maddening reason the visuals look way more appealing. For these reasons I would prefer to use Python

I guess I'm just being biased towards Python as its open source. Each to our own!!

Is there any good webpage comparing Python with Matlab at basic linear algebra times for example?

Python and its companion libraries are getting more and more sophisticated. For example, scikit.learn includes state-of-the-art ML approaches with very good documentation/tutorial/examples. Moreover Python, as a programming language for large scale problems, is lot more expressive and readable as compared to MATLAB scripts. Then why Python is not as popular as MATLAB? May be because (i) it is the legacy of the research community and (b) that most student who start research have diverse backgrounds e.g. Engineering/Bio and they feel comfortable with writing simple MATLAB scripts which after some time becomes a habit.

I think it is just because of the promotion from professors, if they start telling their students: "I want it written in Python", then Matlab and other commercial programms will start vanishing.

So, the hidden question is: "Why do the professors continue using Matlab?"

Matlab Alternatives:

Scripting similar: Octave, Scilab, Freemat

Scripting non-similar: Python

Native languages C, C++, Fortran

https://www.researchgate.net/post/Which_is_the_best_alternative_to_Matlab?exp_tc=tprc

I finally switched from Matlab (using it since version 4.something till finally version 2012a) to Python this year. The difference in consistency in the language and ease of use is impressive - I would never go back if I can (I am still maintaining some opensource Matlab-based tools on sourceforge, so my switch is not complete).

Matlab for the novice user appears easy to learn, it is similar to Fortran and Basic on the first look. Algorithms can easily be converted into matlab code. Plotting is nice and easy, too. So, for the easy things there is no need to change once you are accustomed to Matlab. And most professors probably still have a background in Fortran, so this may explain why Matlab comes as natural and easy choice to them. Also do not forget the thousands of lines of public code written in Matlab.

Things get different in the moment that you want to do more. Get the data out of a database. Do some string manipulation. Try to do real functional or object oriented programming. Have a complex user interface. Big data. All this appeared difficult to do for me in Matlab, but easy in Python. Switch to python and new worlds open up, all of a sudden. :-)

But Python has its down sides also. Numpy and matplotlib are not inlcuded, so you have to install them. This, especially on a Mac, was not as easy as it is today just some years ago. I remember having to install Sage because it was the only reliable method to get Numpy and matplotlib working on the Mac... People with this experience may not realise that today all you need is pip or easy_install, or may fear that from one system update to another Python will fail again.

Another thing is dependecy problem with python. Although there are various useful libraries written in python such as NumPy, SciPy, matplotlib etc., they are not upgraded together. Thus you may not be able to use an upgraded python version or library due to the fact that a library may depend on older version of python or another library.

One rather simple reason to stay with Matlab for me: a lot of the companies that hire my students ask for Matlab / Simulink knowledge / experience. A company like Volkswagen does not really care that Python can be obtained rather inexpensively. They simply ask the following: can Python do all the things that Matlab / Simulink does (e.g.: generate C-Code from Simulink models) and if it can do all that: how much will it cost to move the huge volumes of legacy code to Python. The world of IT shows a high level inertia sometimes....

Why not R?

From my experience, I would say the two main reasons (at least in a research setting) are: 1) most universities tend to have site licenses meaning the fact python is open source offers no advantage, 2) nowadays Matlab tends to be taught at undergrad level for general data processing, so people are more likely to have experience with this than python when they start their career.

I have switched to python for general plotting, since for the same amount of effort the matplotlib library offers far better figures with more options for customising etc. However, for image analysis (filtering, line profiles, edge detection etc) I still use Matlab. I tried migrating this to python and found it tough going - I first tried to generate an image from a numpy array and save it to PNG format. For something so simple, this turned out to be very problematic and frustrating, and took a good two hours of googling, uninstalling/installing different libraries and versions of python.

Admittedly I haven't tried much with python and will probably return to it at some stage - from the comments above it certainly seems to offer more than Matlab. However, I haven't come across anything in terms of image processing that I can't do in Matlab - so long as I have the university license I would say I have no reason to switch to python.

I think the fact is that most people are already accustomed to Matlab by the time they start their career. Transferring those skills to python is something that takes time, and it's well known that the further into your career you go, the less time you have for these sorts of things! Furthermore, installing libraries and getting them to work in the way that you want also takes time - I imagine to most, the self-contained nature of Matlab appeals to a lot of people, not to mention the wealth of material in worked examples, user defined functions etc that are available on the web.

In conclusion: Matlab seems to be easier to use, and most people already know how to use it and have access to it without paying. Python seems to offer more capabilities but requires some investment of time and effort!

Actually you can easily search for python packages that already contain all the requisite libraries! Spyder for instance!! That saves a lot of trouble!!

I think Python (with additional libraries such as scipy, scikit, numpy) can be a good alternative to MATLAB. Especially if you also do lots of text processing besides your machine learning/data mining related code. Text processing in MTALB is rather difficult and you might have to do some preprocessing in some other language (such as Python) before you actually use MATLAB. This might be fine if you do batch processing where you can generate all your features before hand but can be problematic if you have to generate your features dynamically directly from text and then apply whatever ML on those features.

Since we are all scientists talking about scientific python, don't forget that you can get all of these tools in a nice distribution, with a package manager and editor for free. Saves a lot of hassle over building all of the core libraries from source:

https://www.enthought.com/products/canopy/

My impression is that core Matlab use is dying except for some niche areas of optimization (where Matlab excels) or some academic groups run by an older generation. Have a look at Kaggle. Almost all the winners are using R or Python except perhaps for the few contests sponsored by Mathworks that are tailor made to hit Matlab's sweet spots. I don't think there is anything like IPython notebook or pandas available through Matlab. R has similar packages.

Here you can find an interesting report

"Comparative Evaluation of Matlab, Octave, Scilab and others"

http://userpages.umbc.edu/~gobbert/papers/ComanHPCF2012.pdf

I resently switch to PYTHON from MATLAB. So far it is looking good. At first it can be quite consuming if you have to translate MATLAB routines from previous work but I like the flexibility and extra nice option of handling different types of data, lists, arrays and tuples. I think the answer to your question is that MATLAB has been longer than python as a scientific computing, but i think now PYTHON is at the same level.

I may have come a little late to the discussions but at least so far there are some things that I haven´t found how to do them (easily) in Python (wich I love). Take for example the simulation of control systems. Altough you can program the math and transfer functions, and have solver available in scipy, I still think that using Simulink is easier.

Three years later and my impression is that Matlab has lost big part of its share in the research world. Most sci python libraries have now been ported to python 3. IDEs like pycharm, iphyton notebook, jupyter notebook and distributions like anaconda make python far more usable for researchers than years ago. People I know who still use matlab do it because of legacy software or Simulink. Python based notebooks are now even basic tools for the open research (i.e. open data and open methods) movement.

Hi Ardara Kannan!

Let me answer your question as I think I have requisite experience for the same. I was an average MATLAB user till I guess 3 years back,Back then I had an old laptop with windows in it.One fine day I decided to install linux and with that started my python love. What began as a necessity turned into a long lasting relationship. Here are my two cents on this whole issue.

Usage of python can be broadly divided into two realms(thats what I call it)-

1)Simulations

2)Hardware

#1:

Python has a lot of libraries--for almost any kind of simulation I have faced till now. Even if there isnt the kind of library you want--you can always search github repos(someone would have done it) or you can go for writing your own classes and libraries.Believe me writing your own classes in python is way more easy than doing the same in native C or C++!

Since I am a Control Engineer,I will briefly explain you the kind of libraries that I have personally dealt with in python to get my job done-

1)Numpy,Sympy and Scipy for normal scientific computation tasks.

2)PyDStool for Bifurcation Analysis and Continuation ALgorithms

3)Ipopt wrapper for Optimal Control

4)Siconos for hybrid systems.

5)matplotlib for plotting--and believe me it rivals MATLAB!

6)Then there is a library that lets you plot sensor data in real time in python!Amazing isnt it?-"drawnow"

In fact I am currently developing one library for Sliding Mode Control as a pet project of mine.

#2:

Python works well on hardware as well.By hardware I mean small portable micro-computers like Raspberry pi,beaglebone,Intel Galileo etc(I own most of them btw!). You can develop your control algorithm and simulate it using #1 explained above and deploy it directly on any of the boards mentioned above with least possible hassle. Now of course python cant match the raw C performance when it comes to toggling IO pins or Pulse Width Modulation,however the difference isnt perceptible in most cases.One can clearly get away with the difference in speed.

You can also use ESP8266 NodeMCU with micropython(if speed and cost was the only criteria) however ,it will limit the full fledged libraries mentioned in #1(you would instead have to deal with a much reduced library set of python geared towards increasing the performance).

There are tons of libraries for hardware manipulation like MRAA,UPM etc for Intel  Galileo. In fact there is a library for MQTT protocol as well. Just install mosquitto and paho-mqtt in your linux machine and you're ready to rock!

So In a nutshell,the legacy approach would be to

(A) Simulate the control algorithms in MATLAB,

(B)Have MATLAB to generate native C code for target hardware.

However Python cuts these two steps into one--Now you can develop your algorithms and simulate in python and also deploy it directly in hardware of your choice. The speed however would depend on a lot of things--In most cases you can surely get away with it!

As for the reason why it isn't used as much as matlab--it is primarily because of simulink! People hate to code these days, instead they just want nice graphical stuff with minimum hassle--however it costs a lot. A lot of researchers just want to get the job done instead of going deeper into coding aspect.