Table 4.

Prediction errors of contrast sensitivity models, tested on the same subset of data points. The errors are reported in the dB units (see Eq. (33)) as the mean and standard deviation across five-folds (leave-one-out cross-validation). Each column shows the result for the subset of the dataset, containing variation along selected parameters of the stimuli (selected to match the parameters supported by CSFs). The value of N is the number of data points included in each subset (testing and training parts). The predictions are reported for five-fold cross-validation (see text). The symbols denote: ρ, spatial frequency; Y, luminance; a, area; ω, temporal frequency; e, eccentricity, c, modulation along chromatic color directions., c_DKL, modulation along the cardinal color direction (achromatic, red-green, yellow-violet) only. Asterisks (*) mark the cases where low spatial and temporal frequencies are removed from the datasets.

The bold text indicates the model with the minimum prediction error for that particular parameter set within the column.