Research‎ > ‎Distance Learning‎ > ‎

Self-Organizing Map Demo


This demo visualizes the learning process in a self-organizing map. The network is trained with two-dimensional patterns taken randomly from a uniform distribution and scaled appropriately to fit the screen. The neurons start with small random values and are gradually adjusted in oder for the grid to unfold and form a map that represents the whole training dataset. In the ideal convergence the unfolding grid should overlap with the underlying grid depicted in red color.