Thread by @Mohsen_Omrani

Last week I asked Twitter:
“I am fairly good at Matlab coding but not a pro. I would like to get into big data and machine learning projects. Do I really need to switch to Python. Please answer from a practical viewpoint, not an anti-matlab one.” /1

And many of you kindly replied with so many good reasons and resources. So I thought I summarize what I learnt here for people with similar question in mind. /2

The consensus was that “Yes, you really need to switch.” Here are some of the reasons:
- Python is the language for most important autograd tools (PyTorch, TensorFlow, JAX, Keras). /3

- Python is a fully featured language that is used throughout the world in a huge number of domains and is not proprietary.
- Python is free, but MatLab is very expensive, especially outside academia and considering added libraries for more specialized works. /4

- Matlab won’t serve you best in ML/data science because the community is using other tools. Python has many ML packages to deep dive in some ML projects while Matlab ML capacity is still in its infancy. If you want to use DNN, Python is your best choice by a large margin. /5

- There are so many more tutorials, github repos, and amazing libraries to learn from for Python as compared to Matlab. /6

- End of the day Matlab and Python are quite similar and transition is quite smooth and will likely only take a few days.
-There are tools to help: there is an engine in python that enables you to run matlab scripts so you can always pass data sets (as csv) between languages. /7

- It is helpful to test your most critical code by implementing it both in Matlab and python to make sure you are not making a big mistake through transition.

- Here is a quick cheat sheet for switching from Matlab to Python: mathesaurus.sourceforge.net/matlab-numpy.h… /8

- The best way to do the transition is to dive in and do a project. However, the thing many find most different about python is the data structures (dicts, tuples, lists, etc.) That’s the one thing they suggest you read about and understand before you get started. /9

Some other useful resources suggested through the thread were:
@ProjectJupyter @JuliaLanguage @paperswithcode /10

@ProjectJupyter @JuliaLanguage @paperswithcode Big shout out to whomever took the time to give feedback:
@SussilloDavid @_esin_yavuz @SoloGen @JCashaback
@RaghuNC @DrJamieW @yigitdemirag @_3lettersthen7 @andronovhopf @LBurg @chethan @GhorbaniAmirata
@RezaOmrani1 @viuvinhos @john_c_mitchell @chrisgbedford @lukesjulson /end