Cluster Optimized Proximity Scaling

Proximity scaling methods such as multidimensional represent objects in a low-dimensional configuration so that fitted object distances optimally approximate proximities. Besides finding the optimal configuration, an additional goal may be to make statements about cluster arrangement of objects. This fails if lacks appreciable clusteredness. We present optimized proximity (COPS), which attempts find exhibits In COPS, flexible parameterized loss function emphasize differentiation information proximities is augmented with index (OPTICS Cordillera) penalizes lack clusteredness configuration. two variants this, one for directly and hyperparameter selection parametric stresses. apply both functional magnetic resonance imaging dataset on neural representations mental states social cognition task show COPS improves enabling visual identification clusters states. Online supplementary materials are available including R package document details.