confused with the conv layer

[tc-yue]

Att_v = tf.contrib.layers.conv2d(G, num_outputs=opt.num_class,kernel_size=[opt.ngram], padding='SAME',activation_fn=tf.nn.relu) #b * s * c
The implementation code above is a conv2D operation on the match matrix G.
while the formulations in the paper below seems only one filter with the size(2r+1) to produce a further match matrix (K*L). I think they are little different. Is it True?

$$u = Relu(GW+b)$$

where W ∈ R2r+1 and b ∈ R K u l ∈ R K

[tangqianyan]

Excuse me, could this code be applied to multi-label learning directly?

[ReactiveCJ]

@tc-yue it is just a conv1d so the size of W is (2r+1) * K * K

[lanjinraol]

@tc-yue Hello, I have same confusion with you. I think it's different between the two operations. In the paper, the filter with size(2r + 1) only captures contextual feature, while in the implementation code the K filters capture contextual features and relationships among the categories. Have you gotten any reasonable explanation?