Covariance-Generalized Matching Component Analysis for Data Fusion and Transfer Learning

The matching component analysis (MCA) transfer learning technique was originally developed as a data augmentation strategy for building large, representative machine learning training sets within a data-limited environment [1]. Specifically, MCA maps a training domain and a testing domain into a low-dimensional, common domain using only a small number of matched train-test image pairs. These maps minimize the expected distance between train-test image pairs within the common domain, subject to an identity matrix covariance constraint and an affine linear structure. The training domain's optimal affine linear transformation - encoded with information from the matched train-test image pairs - is then applied to a large number of unmatched training images, resulting in a large number of common-domain image representations to be used as training inputs. We are interested in extending the MCA application space to the fusion of data acquired from two different modalities.