[MGNLPER-90] LRU Policy instead of FIFO Created: 12/Feb/19  Updated: 14/Jun/19  Resolved: 26/Feb/19

Status: Closed
Project: Periscope
Component/s: None
Affects Version/s: None
Fix Version/s: 1.1

Type: Story Priority: Neutral
Reporter: Simon Lutz Assignee: Cedric Reichenbach
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Relates
relates to MLEARN-11 Configuration to reduce the size of n... Closed
documentation
Template:
Acceptance criteria:
Empty
Task DoD:
[ ]* Doc/release notes changes? Comment present?
[ ]* Downstream builds green?
[ ]* Solution information and context easily available?
[ ]* Tests
[ ]* FixVersion filled and not yet released
[ ]  Architecture Decision Record (ADR)
Release notes required:
Yes
Documentation update required:
Yes
Date of First Response:
Epic Link: Search relevance
Sprint: Foundation 4, Foundation 5
Story Points: 5

 Description   

Currently we drop results from the output label according to a FIFO-policy.
In order to improve the search relevance, we should drop results according to a LRU policy (last recently used).
With this we would still keep the most often used ones even when we'd reduce the labels.

Also see comments on the linked ticket.



 Comments   
Comment by Cedric Reichenbach [ 15/Feb/19 ]

Documentation updates will only be necessary if the NN result ranker "memory" is actually subject to documentation (not sure if it's planned, hence set the tick to be on the safe side).

Comment by Christoph Meier [ 21/May/19 ]

fgrilli, cc slutz ... I guess this has been implemented, right?

Currently docu claims:

Results are removed from networks based on a least-recently-used policy.

That should be enough, right?

Comment by Simon Lutz [ 21/May/19 ]

Yes, this is implemented and to me the documentation is sufficient as it stands now.

Generated at Mon Feb 12 10:28:38 CET 2024 using Jira 9.4.2#940002-sha1:46d1a51de284217efdcb32434eab47a99af2938b.