Skip to content

Latest commit

 

History

History
19 lines (14 loc) · 872 Bytes

File metadata and controls

19 lines (14 loc) · 872 Bytes

Kohonen Neural Network

A Java library

The network topology is defined by the number of neurons. The learning algorithm of the Kohonen Network uses euclidean distance as a metric

  • Step1: Load input vector
  • Step2: Get best matching unit by calculating the minimal distance between a node's weight vector and the input vector
  • Step 3: The winning node learns the data (weight modification)
  • Step 4: Modify the neighbours weights(the learning fades as a gaussian func)
  • Repeat until number of iteration reached (no more learning vectors)

This library is made to communicate with the OMNet++ simulator via shared files. The network learns the data exchanged within a wireless sensor network for classifying. It can be used for other purpose if ùodified a little bit.

Copyright (c) 2016, Asma DHANE