Self Organizing Maps

This piece of code implements the Self Organizing Maps (SOM) algorithm using numpy.

Using the SOM on different data, one is interested in the similarity mask. It can show clusters in data. The darker an area is the further away are the feature vectors of the SOM from each other. By applying SOM.remove_bmu_nodes(threshold) one can also remove interpolating nodes.

Dependencies:

Built against the Python 2.7 on Linux. Python 3 compatibility unknown.
No OS specific code used, untested on other operating systems.

• python-numpy
• Saving images (like the similarity mask) requires Python Imaging Library (PIL)