Skip to main content
Fig. 3 | BioData Mining

Fig. 3

From: Application of an interpretable classification model on Early Folding Residues during protein folding

Fig. 3

Principle of Generalized Matrix Learning Vector Quantization. Graphical depiction of learning with GMLVQ [63, 64]. One or multiple prototypes represent classes: each data point in the data space of dimension N belongs to the class of the prototype with the closest distance d. Prototypes are updated during learning as in LVQ [65]. Additionally, the matrix Ω maps the data space to an embedded data space of dimension M, where mapped distances d′ are optimized. The matrix Λ=Ω′Ω (CCM) represents the impact of each feature on the classification performance

Back to article page