Clustering algorithm for mutually constraining heterogeneous features
Wolfgang Fink, Rebecca Castaño, A. Davies, Eric Mjolsness · 2001
We introduce a general type of optimization algorithm which infers data models relating two different but intertwined types of information about each of a set of objects. A novel clustering problem is solved by formulating an objective function which is optimized. For the optimization of an objective function describing a general classification problem we use a clocked objective function update scheme. As a concrete example we apply the clustering algorithm to geological data (rocks) to infer the spatial as well as mineral relationships within a field geology model. We test the algorithm with synthetic data generated according to a particularly chosen probability distribution function. 1