Compared to other numerical targeting packages/languages (like R or Octave/Scilab), the Python + Numpy mix seems to worth investigation: one can use her OOP skills offered by Python in conjunction with a linear algebra package. Developing neural networks or prototyping algorithms in computer vision are supported by Numpy. As Python seems to be the language of choice for developing or prototyping Machine Learning systems (see for eaxmple the answers from http://www.quora.com/What-are-the-best-open-source-machine-learning-libraries-written-in-Python), Numpy is certainly a “must” for anyone interested in developing such softwares.