Decay Function question

I've begun using the decay function in order to promote more recent results
in our index. In particular I'm using what's documented here:

http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/query-dsl-function-score-query.html

Here's the date example they use (let's assume gaussian slope):

"DECAY_FUNCTION": {
    "FIELD_NAME": {
          "origin": "2013-09-17",
          "scale": "10d",
          "offset": "5d",
          "decay" : 0.5
    }
}

So in trying to visualize this decay, is it safe to make the following
assumptions in terms of associating these input values to a graph:

  • origin is the start (x-value) of the slope, in this case the date
    9/17/2013
  • for all points up to offset, the slope is flat, so there's no negative
    scoring for the 5 days after 9/17/2013
  • the "end" of the slope is scale
  • at the data point 'scale', the slope will have a y-value (on the graph)
    of decay (0.5)

The last point I am not so sure about. Furthermore, I'm unclear as to what
happens for articles outside of scale, so past 10 days. My test show that
they actually get a 'zero' multiplier for their score, so basically their
scores all end up being zero. I was under the impression that they would
get a score multiplier of 0.5. Any help would be appreciated.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/32d518ad-2b63-4f13-a952-0f408722bd79%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Hi Mario,

I added an image which hopefully explains a little better how decay
functions work here:

  • origin is the start (x-value) of the slope, in this case the date 9/17/2013

yes

  • for all points up to offset, the slope is flat, so there's no negative scoring for the 5 days after 9/17/2013

The decay function never returns a negative value. It will always be
between 0 and 1. For each value within +- offset from the defined
origin the decay function will just return 1.

  • the "end" of the slope is scale

No, decay function will decrease further. The scale parameter just
steers how quickly the function approaches 0.

  • at the data point 'scale', the slope will have a y-value (on the
    graph) of decay (0.5)

yes.

My test show that they actually get a 'zero' multiplier for their score, so basically their scores all end up being zero. I was under the impression that they would get a score multiplier of 0.5.

No, the score will decrease further until it reaches 0. If you need
to have documents outside the "scale" range to have a value of 0.5 you
need to define a separate function which then adjusts the score for
these documents as needed.

On Thu, Nov 6, 2014 at 8:07 PM, Marlo Epres mepres@gmail.com wrote:

I've begun using the decay function in order to promote more recent results
in our index. In particular I'm using what's documented here:

Elasticsearch Platform — Find real-time answers at scale | Elastic

Here's the date example they use (let's assume gaussian slope):

"DECAY_FUNCTION": {
    "FIELD_NAME": {
          "origin": "2013-09-17",
          "scale": "10d",
          "offset": "5d",
          "decay" : 0.5
    }
}

So in trying to visualize this decay, is it safe to make the following
assumptions in terms of associating these input values to a graph:

  • origin is the start (x-value) of the slope, in this case the date
    9/17/2013
  • for all points up to offset, the slope is flat, so there's no negative
    scoring for the 5 days after 9/17/2013
  • the "end" of the slope is scale
  • at the data point 'scale', the slope will have a y-value (on the graph) of
    decay (0.5)

The last point I am not so sure about. Furthermore, I'm unclear as to what
happens for articles outside of scale, so past 10 days. My test show that
they actually get a 'zero' multiplier for their score, so basically their
scores all end up being zero. I was under the impression that they would get
a score multiplier of 0.5. Any help would be appreciated.

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/32d518ad-2b63-4f13-a952-0f408722bd79%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CALhJbBimcB2xX49FqyXyXye7ZUrsq3gDPT31gHF0p4%2B3q%2BaXRA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.