I don't like Matlab. It is incredibly bloated and inefficient. It takes almost a minute (I just clocked it at 55 seconds) to start up on my Macbook Pro. Once it's open, it sits there using 150mb of RAM, and if I start using it, it will quickly climb to 200-300mb. For my DSP class, we have a function that analyzes the frequency components of sounds. I tried running it on a full-length (3 minute 44100hz) song. When I loaded the song, the memory use went up to 300mb, however the data is only about 80mb it Matlab format. When I ran the function, the memory use shot up to 2gb, then it stopped with a memory error. I guess it is a 32 bit application, which is limited to 2gb of address space. (There is a 64-bit beta for OS X, but it is not available to people with the student license.) I ran it on a shorter segment of the song, and it used 1.6gb to analyze about a minute of it. Maybe it's just how the function was written, but there's no reason it should be using that much memory. It should be analyzing just 4 milliseconds at a time, or 8 if you count the overlap. When they introduced Matlab, they told us that loops are "highly inefficient" and that it is always better to use vectors. I don't know what could make loops so inefficient, especially since vector operations are eventually converted into loops when they are executed on the CPU. I haven't done any tests to see how much slower loops are, though.
Ignoring all the inefficiencies in Matlab, it is still not a great programing language in my opinion. They do a lot of things differently from most other languages, like starting indexes at 1 instead of 0 (that's all I can think of off the top of my head, besides obvious syntax differences).
I haven't used any of the other math software mentioned here, though, so I can't make any comparisons.