Does it has something wrong in reconstruct loss code?

[AlexHex7]

        masked = Variable(torch.zeros(input.size())).cuda()
        masked[:,v_max_index] = input[:,v_max_index]

The version of python I use is 3.5.
In my opinion, the v_max_index (batchsize, 1) means for each sample of batch, there is a max length vector among 16 vectors. So for the sample 0, there should be active vector among 16 vectors, and other 15 vectors are all 0.
But for

masked[:,v_max_index] = input[:,v_max_index]

it means each sample, there are v_max_index vector will be assign which includes duplicate assignments.

And the paper says

During training, we mask out all but the activity vector of the correct digit capsule.

The figure 2 in the paper show that for each sample, there are 15 vectors among 16 vectors will be masked.

[timomernick]

Oh I just found what you mean. It's duplicating the assignments over the batch dimension. I'll fix it, thanks for pointing this out!

[timomernick]

To verify, here is what the masked capsule activity looks like after my fix. Only one capsule is nonzero for each batch sample. Here there is a batch size of 2. The first sample has capsule 0 active, and the second sample has capsule 2 active.

         [ -3.17241793e-04],
         [ -7.09305750e-05],
         [  2.88858573e-05]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]]],


       [[[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[ -3.76116368e-04],
         [  2.83236353e-04],
         [  1.13728966e-04],
         [ -8.82451714e-05]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]],

        [[  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00],
         [  0.00000000e+00]]]], dtype=float32))