[
https://issues.apache.org/jira/browse/SOLR-11597?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16353937#comment-16353937
]
Michael A. Alcorn commented on SOLR-11597:
------------------------------------------
Hi, [~cpoerschke]. Please see my latest commit
[here|https://github.com/apache/lucene-solr/pull/270/commits/96746d1e97380848d06eaea03d52a892af6f3794].
{quote}So I think we can stick with the matrix approach here...{quote}
Based on this feedback and the rest of the discussion in this Jira, the model
representation now takes the following form:
{code:json}
"layers" : [
{
"matrix" : [ [ 1.0, 2.0, 3.0 ],
[ 4.0, 5.0, 6.0 ],
[ 7.0, 8.0, 9.0 ],
[ 10.0, 11.0, 12.0 ] ],
"bias" : [ 13.0, 14.0, 15.0, 16.0 ],
"activation" : "relu"
},
{
"matrix" : [ [ 17.0, 18.0, 19.0, 20.0 ],
[ 21.0, 22.0, 23.0, 24.0 ] ],
"bias" : [ 25.0, 26.0 ],
"activation" : "relu"
},
{
"matrix" : [ [ 27.0, 28.0 ] ],
"bias" : [ 29.0 ],
"activation" : "none"
}
]
{code}
I've also added a {{Layer}} class that I believe makes the logic of the
calculations clearer.
bq. On the non-linearity vs. activation-function choice, let's go with
activation or activation function.
Done ("activation").
bq. And can I suggest we take a function/functor style approach...
Yep, agreed that's a better approach.
bq. Might it be helpful to include the input values somehow...
Added them to {{explain}}.
bq. Using StringBuilder in explain...
Done.
> Implement RankNet.
> ------------------
>
> Key: SOLR-11597
> URL: https://issues.apache.org/jira/browse/SOLR-11597
> Project: Solr
> Issue Type: New Feature
> Security Level: Public(Default Security Level. Issues are Public)
> Components: contrib - LTR
> Reporter: Michael A. Alcorn
> Assignee: Christine Poerschke
> Priority: Major
> Fix For: 7.3
>
>
> Implement RankNet as described in [this
> tutorial|https://github.com/airalcorn2/Solr-LTR].
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
