A&C_IADI Logo
HOME ABOUT VOLUMES SEARCH

Volumes / Supplements / 6 - 2014 / Manni F., Coppa A., Candilio F.

Kohonen self-organizing Maps to unravel patterns of dental morphology in space and time

Click on any image to open slideshow

Fig. 1 – ASUDAS scoring system (Turner et al. 1991; Scott, Turner 1997) for the description of human dental morphology. Example of a dental trait scored on lower molars (the Protostylid) that can be either present (M1) or absent (M2). Our analyses were made according to 23 traits of this kind, thus giving rise to vectors having 23 binary descriptors.

Fig. 1 – ASUDAS scoring system (Turner et al. 1991; Scott, Turner 1997) for the ...

Fig. 2 – Examples of data vectors defined by eight dental traits (A; B; C; D; E; F; G; H). On the top of the figure a dataset of 6 individuals is displayed and the descriptors correspond to the presence (1) or absence (0) of a given dental trait. Missing measures are reported as “?”. As all vectors present missing descriptors, we visually suggest the various strategies that are generally adopted to analyze them. In analytical projections (like MDS or PCA), no missing data can be processed, therefore individual samples can be converted in a population vector whose descriptors correspond to a majority rule consensus concerning available descriptors (case A). Otherwise, a compromise between the number of individuals or the number of observations that are kept for the analysis has to be achieved. In B there are 4 individual vectors (#2; #3; #5; #6) with 4 traits (B; C; E; G; H); in C there are less individuals (#2; #3; #6) with more descriptors defining them (B; C; E; G; H). Artificial Neural Network analysis can be used to process the full set. Please note that the vector corresponding to “Individual #1” has been excluded from analyses due to its too many missing descriptors.

Fig. 2 – Examples of data vectors defined by eight dental traits (A; B; C; D; E;...

Fig. 3 – Example of a Self-Organizing Map (Kohonen 1982) concerning the analysis of 126 dentitions from different geographic locations and periods. The map is a 12×12 lattice corresponding to 144 neurons. Several samples can be mapped to a same neuron (not shown). We display only the position of the sample Tabun BC7 that falls in an area whose neurons became specialized in the description of the Neanderthal teeth morphotype. The map has been colored according to time periods and geographical areas. Gray cells correspond to empty neurons, meaning that no inputs were linked to them in the classification process (redrawn from Coppa et al. 2007a).

Fig. 3 – Example of a Self-Organizing Map (Kohonen 1982) concerning the analysis...

Manni F., Coppa A., Candilio F. 2014, Kohonen self-organizing Maps to unravel patterns of dental morphology in space and time, Archeologia e Calcolatori 6