I deeply regret writing my PhD software in Matlab.

From 2009-2013 I was stuck struggling with the Matlab for my thesis. It made sense at the time, Matlab was the most advanced language I was taught in undergrad, it had amazing linear algebra capabilities, it was fast but not as difficult as C++. Yet once my code became mildly complex things fell off the rails.

Here are 3 reasons I would say to advise new PhD students to not use MATLAB.

Matlab is not free

When you're a student Matlab costs about the price of a textbook.

Once your in industry, Matlab quickly becomes a liability. During my first Post-Doc job if I wanted access to Matlab I had to share access to a pool of 20 licences shared across 1000 researchers. I could either be unproductive or try and fight for access to a suitable licence.

It's far easier to use free software like Python instead.

Software Skills Are Valuable For Researchers

Students often say "but I don't want to learn software, so I'll use MATLAB".

It's not that much harder to learn to do things that MATLAB does but in Python. Libaries such as Pandas, Numpy and Pytorch give you access to scientific computation you need, but for free.

If you ever decide to change careers, your years of Python experience will be looked upon favourably.

Deployment and sharing is much easier

There are now huge opportunities for deployment using Python.

You could use a Google Colab notebook, or a Kaggle notebook or pay for AWS Sagemaker. All of these free or nearly free services will happily run and host your software. Allowing you to get your engineering and software ideas out there.

The community of python developers is also so much larger, which means you can easily share and collaborate with many more people.