You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@predictionio.apache.org by Vaghawan Ojha <va...@gmail.com> on 2017/04/07 08:02:14 UTC
How Does The Score Works In UR
Hi,
I was just curious, how does the score works in the UR, I've seen the UR
presentation in the documentation site, but I couldn't be sure what it
really signifies.
For example: what's the limit of the score? What's the possibly highest
score?
Thanks
Re: How Does The Score Works In UR
Posted by Vaghawan Ojha <va...@gmail.com>.
Hi Pat,
I did some study and understood some of those complexities, and some are
yet far to be understood. You have put really nice presentation in the doc
site.
I really appreciate your work and help.
Thank you very much!
On Sat, Apr 8, 2017 at 10:12 AM, Vaghawan Ojha <va...@gmail.com>
wrote:
> Hi Pat,
>
> Thank you very much for the explanation, seems like I forgot to properly
> search this question in the group. By ranking you means the order of the
> product right?
>
> Also I wanted to replicate such dot products of all user history vector
> segments against the correlator fields in the model in a very very small
> data, so that I could have a test of what's really working behind. And also
> I wanted to learn about the implicating algorithms. I've seen the
> presentation in the doc site of UR. Should you have any resources which
> could lead me to better understand all these please point me.
>
> Thanks
> Vaghawan
>
> On Sat, Apr 8, 2017 at 12:16 AM, Pat Ferrel <pa...@occamsmachete.com> wrote:
>
>> The limit depends on your data. I tell people to ignore the value as you
>> do when searching (no one uses the score when doing searches do they?) Only
>> the ranking/ordering makes any real difference.
>>
>> The score is the sum of dot products of all user history vector segments
>> against the correlator fields in the model. If you boost or add properties
>> to your queries they multiply times some of the dot products before
>> summation. So given any boosting the max is hard to state and we have found
>> no particular use for it. Other tuning parameters give you the best ways of
>> affecting the score so looking at the score afterwards is not very helpful
>> except to rank results.
>>
>>
>> On Apr 7, 2017, at 1:02 AM, Vaghawan Ojha <va...@gmail.com> wrote:
>>
>> Hi,
>>
>> I was just curious, how does the score works in the UR, I've seen the UR
>> presentation in the documentation site, but I couldn't be sure what it
>> really signifies.
>>
>> For example: what's the limit of the score? What's the possibly highest
>> score?
>>
>> Thanks
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "actionml-user" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to actionml-user+unsubscribe@googlegroups.com.
>> To post to this group, send email to actionml-user@googlegroups.com.
>> To view this discussion on the web visit https://groups.google.com/d/ms
>> gid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahzMeR86c5MP
>> 6EL%2BCr32ZzA%40mail.gmail.com
>> <https://groups.google.com/d/msgid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahzMeR86c5MP6EL%2BCr32ZzA%40mail.gmail.com?utm_medium=email&utm_source=footer>
>> .
>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>
Re: How Does The Score Works In UR
Posted by Vaghawan Ojha <va...@gmail.com>.
Hi Pat,
Thank you very much for the explanation, seems like I forgot to properly
search this question in the group. By ranking you means the order of the
product right?
Also I wanted to replicate such dot products of all user history vector
segments against the correlator fields in the model in a very very small
data, so that I could have a test of what's really working behind. And also
I wanted to learn about the implicating algorithms. I've seen the
presentation in the doc site of UR. Should you have any resources which
could lead me to better understand all these please point me.
Thanks
Vaghawan
On Sat, Apr 8, 2017 at 12:16 AM, Pat Ferrel <pa...@occamsmachete.com> wrote:
> The limit depends on your data. I tell people to ignore the value as you
> do when searching (no one uses the score when doing searches do they?) Only
> the ranking/ordering makes any real difference.
>
> The score is the sum of dot products of all user history vector segments
> against the correlator fields in the model. If you boost or add properties
> to your queries they multiply times some of the dot products before
> summation. So given any boosting the max is hard to state and we have found
> no particular use for it. Other tuning parameters give you the best ways of
> affecting the score so looking at the score afterwards is not very helpful
> except to rank results.
>
>
> On Apr 7, 2017, at 1:02 AM, Vaghawan Ojha <va...@gmail.com> wrote:
>
> Hi,
>
> I was just curious, how does the score works in the UR, I've seen the UR
> presentation in the documentation site, but I couldn't be sure what it
> really signifies.
>
> For example: what's the limit of the score? What's the possibly highest
> score?
>
> Thanks
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "actionml-user" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to actionml-user+unsubscribe@googlegroups.com.
> To post to this group, send email to actionml-user@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/
> msgid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahz
> MeR86c5MP6EL%2BCr32ZzA%40mail.gmail.com
> <https://groups.google.com/d/msgid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahzMeR86c5MP6EL%2BCr32ZzA%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
> For more options, visit https://groups.google.com/d/optout.
>
>
Re: How Does The Score Works In UR
Posted by Pat Ferrel <pa...@occamsmachete.com>.
The limit depends on your data. I tell people to ignore the value as you do when searching (no one uses the score when doing searches do they?) Only the ranking/ordering makes any real difference.
The score is the sum of dot products of all user history vector segments against the correlator fields in the model. If you boost or add properties to your queries they multiply times some of the dot products before summation. So given any boosting the max is hard to state and we have found no particular use for it. Other tuning parameters give you the best ways of affecting the score so looking at the score afterwards is not very helpful except to rank results.
On Apr 7, 2017, at 1:02 AM, Vaghawan Ojha <va...@gmail.com> wrote:
Hi,
I was just curious, how does the score works in the UR, I've seen the UR presentation in the documentation site, but I couldn't be sure what it really signifies.
For example: what's the limit of the score? What's the possibly highest score?
Thanks
--
You received this message because you are subscribed to the Google Groups "actionml-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to actionml-user+unsubscribe@googlegroups.com <ma...@googlegroups.com>.
To post to this group, send email to actionml-user@googlegroups.com <ma...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahzMeR86c5MP6EL%2BCr32ZzA%40mail.gmail.com <https://groups.google.com/d/msgid/actionml-user/CA%2B69RXbuZU6ujHgQWpTizTigXpvahzMeR86c5MP6EL%2BCr32ZzA%40mail.gmail.com?utm_medium=email&utm_source=footer>.
For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.