Potentialities of neural networks (2)

Efficient representation of multimedia data can be achieved by using unsupervised neural networks, such as the self-organizing feature map (SOFM) and the principal component neural networks.

The Expectation Maximization (EM) algorithm is a well-established iterative method for maximum likelihood estimation and for clustering a mixture of Gaussian distributions. Radial basis function (RBF) NN constitute a neural implementation of the EM algorithm.

Independent component analysis minimizes the output kyrtosis and it can be seen as a generalization of PCA.

Efficient representation of multimedia data can be achieved by using unsupervised neural networks, such as the self-organizing feature map (SOFM) and the principal component neural networks.