English to French Science Computers: Systems, Networks Mapping algorithm

This has to do withh calculations of probabilities for cells in a map, based on the following theory:
In this paper, we investigate the use of maximum variance unfolding (MVU) [24] for the simultane-ous dimensionality reduction of data from different input modalities. Though the original algorithm does not solve this problem, we show that it can be adapted to provide a compelling solution. In its original formulation, MVU computes a low dimensional embedding that maximizes the variance of its outputs, subject to constraints that preserve local distances. We explore a modification of MVU that computes a joint embedding of high dimensional inputs from different data sources. In this joint embedding, our goal is to discover a common low dimensional representation of just those degrees of variability that are correlated across different modalities. To achieve this goal, we design the embedding to maximize the inter-source correlation between aligned outputs while preserving
the local, intra-source distances. By analogy to MVU, we call our approach maximum covariance unfolding (MCU).