Generally, unsupervised learning or self-organized learning finds regularities in the data represented by the examples. Clustering methods such as model-based, density based and user guided methods are often applied for data reduction such as summarization like preprocessing of classification; compression like vector quantization; and finding the nearest neighbors. Specifically, a feed-forward neural network is a software version of the human brain and have their roots in Hebbian and competitive learning such as Kohonen’s self-organizing map and growing neural gas. In this network, data processing has only one forward direction from the input layer to the output layer without any cycle or backward movement; and generally exhibits several advantages such as an inherent distributed parallel processing architectures, as well as capabilities to adjust the interconnection weights to learn and describe suitable clusters, process vector quantization prototypes and distribute similar data without class labels to describe the clusters, control noisy data, cluster unknown data, and learn the types of input values on the basis of their weights and properties. The current online dynamic unsupervised feed-forward neural network clustering methods such as evolving self-organizing map and dynamic self-organizing map inherit some of the advantages and disadvantages of static unsupervised feed-forward neural network clustering methods; which are suitable to be applied in different research areas such as email logs, networks, credit card transactions, astronomy and satellite communications. Generally, the critical issues of clustering are data losing, definition of clustering principles, number and Unsupervised clustering is a valuable subject to research, however, their critical issues are data losing, definition of clustering principles, number and densities of clusters. Specially, the main problems in dynamic feed-forward neural network clustering are low speed, high memory usage and memory complexity through using random weights and parameters, and relearning. The goal of this research is an investigation of current unsupervised clustering and identify their limitations and problems through a literature review and experience.

Aims and Scope:

The topics of the special session include, but are not limited to:
Learning and Neural Network
Unsupervised Feed Forward Neural Network clustering
Static Unsupervised Neural Network clustering
Dynamic Unsupervised Neural Network clustering
Semi-supervised Neural Network clustering