Category_TCGrads.txt (Fig 2): Once unpickled, this file contains a list, which itself contains 2 lists: the first corresponding to tuning curves and the second gradient values. Within these lists are 13 numpy arrays (one for each layer). These arrays have dimension 20 (the number of categories) by the # of feature maps at that layer.

Category_Perfs.txt (Figs 3 and 4): Once unpickled, this file contains a list which itself contains two lists: the first for merged images and the second for array. These lists each contain two numpy arrays, the first corresponding to attention applied with tuning curves and the second with gradients. These arrays are of dimensions 2 (true positive and true negative rates) x 20 (categories) x 13 (layers) x 27 (beta values, here ranged from 0 to 4 in increments of .15). 

OriSpatFeat_Perfs.txt (Fig 5E): Once unpickled, this file contains a list which contains 4 numpy arrays corresponding to attention applied to both spatial & feature attention, spatial attention only, feature attention alone (tuning) and feature attention alone (gradients). Each of these arrays has dimension 2 (true positive and false positive rates) x 13 (layer) x 8 (orientation) x 6 (beta values).

Ori_Perfs.txt (Fig 5D): Once unpickled, this file contains a list which contains 2 numpy arrays corresponding to attention applied to tuning or to gradients. Each of these arrays has dimension 2 (true positive and true negative rates) x 13 (layer) x 8 (orientation) x 12 (beta values from 0 to 2.2 in steps of .2).

Inter_Slopes_Grad.txt (Fig 6C,right): Once unpickled, this file contains a list which itself contains 5 lists, representing the 5 layers at which attention was applied in Fig 6 (according to gradients). In these lists are 13 numpy arrays that are each 2 x # of feature maps (with Nans removed). The first dimension is the intercept and the second the slope of the fitted line of activity ratios.

Inter_Slopes_TC.txt (Fig 6C,left): Once unpickled, this file contains a list which itself contains 5 lists, representing the 5 layers at which attention was applied in Fig 6 (according to tuning). In these lists are 13 numpy arrays that are each 2 x # of feature maps (with Nans removed). The first dimension is the intercept and the second the slope of the fitted line of activity ratios.

Ori_TCGrads.txt (Fig 7): Once unpickled, this file contains a list, which itself contains 3 lists: the first corresponding to tuning curves, the second detection gradient values and the third to color classification gradients. Within these lists are 13 numpy arrays (one for each layer). These arrays have dimension 9 (the number of categories) by the # of feature maps at that layer.





 


