Part of Advances in Neural Information Processing Systems 11 (NIPS 1998)
Te-Won Lee, Michael Lewicki, Terrence J. Sejnowski
We present an unsupervised classification algorithm based on an ICA mixture model. The ICA mixture model assumes that the observed data can be categorized into several mutually exclusive data classes in which the components in each class are generated by a linear mixture of independent sources. The algorithm finds the independent sources, the mixing matrix for each class and also computes the class membership probability for each data point. This approach extends the Gaussian mixture model so that the classes can have non-Gaussian structure. We demonstrate that this method can learn efficient codes to represent images of natural scenes and text. The learned classes of basis functions yield a better approximation of the underlying distributions of the data, and thus can provide greater coding efficiency. We believe that this method is well suited to modeling structure in high-dimensional data and has many potential applications.