Hi, we're looking at using this package in some upcoming work so I'm just trying to understand the implementation a little better. Currently, I'm hung up on how you've implemented the hub-penalization as described in equation 2 of Akhmedov 2017.
p'_{v} = p_{v} - mu x degree(v)
From my understanding of the linked interpretation this is only penalizing the prizes of steiner nodes and not those given prizes, and conversely only the prize-nodes are scaled by b which isn't how I'd interpreted the equation in your manuscript.
Although, I see having just checked Forest's source code they do the same thing so I guess it was a misinterpretation on my behalf. Regardless, I can see value in prize-nodes being penalized by their degree as well. Is this something you ever considered doing or tested at all?
Thanks,
Sam
|
# Update the node prizes |
|
node_prizes = b*node_prz |
|
index = which(node_prizes==0) |
|
node_prizes[index] = hub_penalization[index] |
Hi, we're looking at using this package in some upcoming work so I'm just trying to understand the implementation a little better. Currently, I'm hung up on how you've implemented the hub-penalization as described in equation 2 of Akhmedov 2017.
p'_{v} = p_{v} - mu x degree(v)From my understanding of the linked interpretation this is only penalizing the prizes of steiner nodes and not those given prizes, and conversely only the prize-nodes are scaled by
bwhich isn't how I'd interpreted the equation in your manuscript.Although, I see having just checked Forest's source code they do the same thing so I guess it was a misinterpretation on my behalf. Regardless, I can see value in prize-nodes being penalized by their degree as well. Is this something you ever considered doing or tested at all?
Thanks,
Sam
PCSF/R/PCSF.R
Lines 127 to 130 in 4e5f270