Question regarding sample numbers in layer 1 and 2 about graphsage HOT 3 CLOSED

Hi, I think according to the code here The number is applied in reversed order. So sample_2 should correspond to the number 1-hop neighbors, and sample_1 should correspond to the number 2-hop neighbors.

from graphsage.

Yes that's correct. Thanks for the explanation. The layer here means GCN layer, which is in reverse order of hops.

from graphsage.

Hi, I am still a little confused.
suppose s1 = 25, s2 = 10. In this case s1 != s2.
so, h0 = hidden in L299, h1 = hidden_next after L321 loop.
What confuses me is that I guess h1[0] is aggregated on 10 neighbors, while h1[1] is aggregated on 25 neighbors. Should these be the same? Or 1-hop nodes always aggregated on 25(s1) 2-hop nodes, while target nodes aggregated on 10(s2) 1-hop nodes?
Likely, suppose K=3, s1 = 7, s2 = 6, s3 = 5. Should the target node always aggregate on 5(s3) 1-hop nodes?
Although in practice I guess there is little difference if s_i is similar with each other.

from graphsage.