Re: Assignment 6
6 views
Skip to first unread message
Ni Lao
Mar 21, 2012, 8:35:53 PM3/21/12
to Freddy Chua, machine-learning-with-large-d...@googlegroups.com
I guess if there is no outlink for a node, we can just treat its
corresponding row in P as all zeros.
Ni
corresponding row in P as all zeros.
Ni
On Wed, Mar 21, 2012 at 7:32 PM, Freddy Chua <fre...@cmu.edu> wrote:
> Hi Ni,
>
> I spoke to you earlier about the degree of a node, some nodes have no out
> degree so it resulted in a divide by zero error.
>
> Freddy
Freddy Chua
Mar 24, 2012, 12:18:33 AM3/24/12
to Ni Lao, machine-learning-with-large-d...@googlegroups.com
Some nodes link to another node multiple times. I wonder how do we handle that? Should we treat it as 3 separate links or a single link?
-Freddy
Ni Lao
Mar 24, 2012, 10:34:37 AM3/24/12
to machine-learning-with-large-d...@googlegroups.com
both would be fine, but I guess it is simpler to treat them as separate.
Ni
Ni
Freddy Chua
Mar 25, 2012, 11:18:31 AM3/25/12
to machine-learning-with-large-d...@googlegroups.com, Ni Lao (cs.cmu)
About this statement, "Finally, take the nodes that have non-zero weight in p, and include all the edges that are incident on these nodes. Since both p and r are sparse, they can be kept in memory."
Does the incident edges on these nodes refer to their out links or in links or both?
-Freddy
Does the incident edges on these nodes refer to their out links or in links or both?
-Freddy
Ni Lao
Mar 25, 2012, 11:43:13 AM3/25/12
to machine-learning-with-large-d...@googlegroups.com
I guess it means "edges among these nodes"
Ni
Ni
Kai Xia
Mar 25, 2012, 11:51:27 AM3/25/12
to machine-learning-with-large-d...@googlegroups.com
1. And, in calculating volume(S) = sum {d(u)} for each u in S, Does these d happen to be all in edges and out edges of u in the whole graph? Or only the edges in the subset graph?
2. I met some cases the total degree of some u are zero here. how do we deal with them? Assign a small positive number to d(u) and ignore this independent node?
3. the first line of the "test.adj" file is "(\tab)London" without a node name. Do we ignore? Or treat it as a node names empty?
Kai Xia
If you don't live your life on the edge, you are taking too much space ~~~
Ni Lao
Mar 25, 2012, 11:57:32 AM3/25/12
to machine-learning-with-large-d...@googlegroups.com
On Sun, Mar 25, 2012 at 11:51 AM, Kai Xia <kx...@cmu.edu> wrote:
> 1. And, in calculating volume(S) = sum {d(u)} for each u in S, Does these d
> happen to be all in edges and out edges of u in the whole graph? Or only the
> edges in the subset graph?
only in the subgraph> 1. And, in calculating volume(S) = sum {d(u)} for each u in S, Does these d
> happen to be all in edges and out edges of u in the whole graph? Or only the
> edges in the subset graph?
> 2. I met some cases the total degree of some u are zero here. how do we deal
> with them? Assign a small positive number to d(u) and ignore this
> independent node?
"I guess if there is no outlink for a node, we can just treat its
corresponding row in P as all zeros."
> 3. the first line of the "test.adj" file is "(\tab)London" without a node
> name. Do we ignore? Or treat it as a node names empty?
>
as a node names empty
Tarun Sharma
Mar 25, 2012, 12:44:48 PM3/25/12
to machine-learning-with-large-d...@googlegroups.com
Hi Ni
Volume of S is defined as number of edges incident on S. Incident means falling on, so for a directed graph does that mean we only consider the in degree of all nodes in S, since in degree would capture the meaning of incident.
Thanks
--
Tarun Sharma
Tarun Sharma
Graduate Research Assistant
School of Computer Science
Carnegie Mellon University
Ni Lao
Mar 25, 2012, 1:31:45 PM3/25/12
to machine-learning-with-large-d...@googlegroups.com
I guess it is fine as long as |boundary|<= |volume|.
So either count out/in links only or both in and out links.
Ni
Reply all
Reply to author
Forward
0 new messages