Fig. 4

Balanced error rate (BER) of predictions within plant (IntraPlant), across plants from the same accession (IntraGenotype) and across plants from different accessions (InterGenotype) for the different algorithms compared in this study. The solid circles indicate the median of all predictions for a given algorithm and type of comparison. The error bars indicate the interquartile range (i.e. between percentiles 25% and 75%) of these predictions. The crosses beyond the error bars indicate the BER achieved for each algorithm and type of comparison. As K-means clustering is an unsupervised classification method, only IntraPlant predictions are shown for this method. Supervised classification algorithms are sorted according to their median performance across accessions (InterGenotype), from worst to best. Dashed vertical lines represent the median BER across all predictions by all supervised classification algorithms for each type of comparison. Three plants were used for each accession for a total of 15 plants