Figure S5: Classification performance values under the null hypothesis

In order to assess whether a given classification performance value was statistically significant or not, we compared the values against those obtained from the null hypothesis. The null hypothesis states that there is no statistically significant difference across the different object categories. The performance level under the null hypothesis was obtained by randomly shuffling the object categories (n=100 iterations) and repeating the same training/testing procedure used with the real data. Here we show the distribution of classification performance values obtained under the null hypothesis (bin size=0.01). The multiple green vertical dashed lines indicate 1 through 5 standard deviations. Throughout the text, we used 0.57 as the threshold for statistical significance. This value corresponds to a difference of 3 standard deviations from the null hypothesis.


Top