Please be informed that the Numpy+mkl modules comprise Intel MKL, Fortran, C, DAAL, TBB, and OpenBLAS runtime DLLs required by other modules that are arranged with one of those. You can apparently get most modules working without numpy-mkl by installing the runtime DLLs externally.
And I believe that the Intel Python distribution is organized with an Intel compiler instead of MSVC and also substitutes some functions of certain libraries(numpy, scipy, scikit-learn) with their own optimized methods.
And regarding the packages, you should get from your link and which ones should you get from regular pip?
Probably numpy+mkl and scipy. MKL_fft and numexpr are also nice. But, the standard numpy and scipy binaries are being used by OpenBLAS, which is really faster on AMD CPUs.
If you are a total beginner and wish to learn Python, then do check out the below Python tutorial video for better understanding: