Time Adaptive Self-Organizing Map

A Time Adaptive Self-Organizing Map or TASOM is an artificial neural network that has the potential to learn continuously from its environment. The TASOM networks are shown to be able to work for both stationary and nonstationary environments. They can follow even abrupt changes of the environment. A TASOM network is similar to a Self-Organizing Map (SOM) network in that they both have output lattice on which neurons reside, and competitive learning is the main learning algorithm for the weights of neurons. However, the TASOM network has adaptive learning rates and neighborhood functions whereas the SOM network has often decreasing learning rates and shrinking neighborhood sizes. Moreover, the TASOM has another mechanism to deal with rapid changes in size of the environment, which continuously approximates the size of the environment and subsequently employs it in the TASOM algorithm.
Some versions of the TASOM networks have adaptive number of neurons to better deal with non-stationary environments.
In addition, the TASOM algorithm has been shown to have some similarity to artificial immune systems .
The TASOM has been used for numerous applications such as adaptive principal components analysis , segmenting color images , automatic multilevel threshoidlng of gray-level images , adaptive clustering , active contour modeling, shape modelling , and human eye sclera detection and tracking

Main Menu

See Also

Main Menu