GENERATING VALID EUCLIDEAN DISTANCE MATRICES

GENERATING VALID EUCLIDEAN DISTANCE MATRICES

Intro

Molecular dynamics simulation takes too much time sampling.

Generating Euclidean distance matrices (EDM)

It tries to generate a positive semi-definitive matrice, in order to parameterize the problem in a space (like a vector space).

Here we propose to parameterize an arbitrary symmetric matrix, as the set of symmetric matrices behaves like a vector space. The symmetric matrix can be transformed into a symmetric positive semi-definite matrix.

Euclidean distance matrix WGAN

Consider the Wasserstein GANs (WGANs).

Appendix

Rejected, check this website.

However, they raised some concerns about the actual significance of the approach. The AC shares these concerns; the methodology essentially amounts to constraining the output of a neural network to be symmetric and positive semidefinite, which is in turn equivalent to producing a non-negative diagonal matrix (corresponding to the eigenvalues). As a result, the AC recommends rejection, and encourages the authors to include simple baselines in the next iteration.