One project I’m working on at the moment involves exploring a dynamic extension of the Isomap algorithm for visualising constantly varying real-world road networks. Currently, we are testing out the method on a small scale simulated road network, and most of the original code (written by Laurens van der Maaten, with updates by Alexei Pozdnoukhov), was done in Matlab. Since this work is eventually going to have to run on relatively large datasets, and probably behind the scenes on a server somewhere, we decided that Python was the way to go. The goal therefore was to reproduce the Matlab code using only Python libraries, and the fewer additional libraries required, the better.
The most difficult stage in all this was to convert the Matlab code to Python code, while still remaining relatively fast and simple. The solution is of course the NumPy Python library, and nothing could have made this conversion more simple than this pdf document. It is basically a syntax conversion chart between Matlab/Octave, Python, and R… brilliant!
Check out Vidar Bronken Gundersen’s Mathesaurus site for this, and other useful resources for converting between different mathematical computation environments.
