Graph Embedding with Data Uncertainty

spectral-based subspace learning is a common data preprocessing step in many machine pipelines. The main aim to learn meaningful low dimensional embedding of the data. However, most methods do not take into consideration possible measurement inaccuracies or artifacts that can lead with high uncertainty. Thus, directly from raw be misleading and negatively impact accuracy. In this paper, we propose model training using probability distributions; each point represented by Gaussian distribution centered at original having variance modeling its We reformulate Graph Embedding framework make it suitable for distributions study as special cases Linear Discriminant Analysis Marginal Fisher techniques. Furthermore, two schemes uncertainty based on pair-wise distances an unsupervised supervised contexts.