MESS: Manifold Embedding Motivated Super Sampling

Many approaches in the field of machine learning and data analysis rely on assumption that observed lies lower-dimensional manifolds. This has been verified empirically for many real sets. To make use this manifold one generally requires to be locally sampled a certain density such features can observed. However, increasing intrinsic dimensionality set required introduces need very large sets, resulting faces curse dimensionality. combat increased requirement local we propose framework generate virtual points faithful an approximate embedding function underlying observable data.