Detecting Dark Matter Halos using Principal Component Analysis

Principal Component Analysis (PCA)-based pattern recognition is employed to predict the locations of dark matter halos within simulated galaxy distributions. Local maxima in the net tangential force produced by galaxy positions and ellipticities forms a feature space upon which PCA is performed to distinguish halo from non-halo maxima. Within a dimensionally reduced space large halo types are successfully clustered into two types, facilitating smart subtraction from the tangential force landscape in order to better resolve minor halos. The strengths and weaknesses of this novel approach are examined.