This is true. Even the Le Cunn group has switched to PyTorch I think.
It's also because no one can work out how to use matplotlib properly, so you need to look at about 1000 StackOverflow questions per line of code.
It's also because no one can work out how to use matplotlib properly, so you need to look at about 1000 StackOverflow questions per line of code.
and back in 2010 or so, python's numerical optimization libraries left a lot to be desired, and totally hopeless compared to matlab. (edit, i don't remember it being nearly as good as octave, which wasn't particularly good either back then).