Minimum Curvilinearity explanatory example
Let’s reduce for simplicity our feature space to two variables V1 and V2, where two different sample groups are distributed in two clusters (blue and yellow) according to an unknown nonlinear relation. The Euclidean Distance (on the left) is not suitable to define the real topological distance between points A and B: it does not respect the internal topological relation between the data. Whereas, the Minimum Curvilinear (MC) distance (on the right) exploits the MST path to navigate and compute the nonlinear (curvilinear) distance between the points A and B. All the pairwise distances between the points are stored in the MC-distance-matrix also called MC-Kernel.
References
Machine learning and computational biology
Nonlinear dimension reduction and clustering by Minimum Curvilinearity unfold neuropathic pain and tissue embryological classes
CV Cannistraci, T Ravasi, FM Montevecchi, T Ideker, M Alessio
Bioinformatics 26 (18), i531-i539, 2010
Minimum curvilinearity to enhance topological prediction of protein interactions by network embedding
CV Cannistraci, G Alanis-Lobato, T Ravasi
Bioinformatics 29 (13), i199-i209, 2013
Machine learning and network science
Machine learning meets complex networks via coalescent embedding in the hyperbolic space
Alessandro Muscoloni, Josephine Maria Thomas, Sara Ciucci, Ginestra Bianconi & Carlo Vittorio Cannistraci
Nature Communicationsvolume 8, Article number: 1615 (2017)
Minimum curvilinear automata with similarity attachment for network embedding and link prediction in the hyperbolic space
A Muscoloni, CV Cannistraci
arXiv preprint arXiv:1802.01183, 2018