#101 closed feature request (rejected)
Caching scheme to boost performance of GP evaluation
Reported by: | gkronber | Owned by: | gkronber |
---|---|---|---|
Priority: | low | Milestone: | HeuristicLab 3.3.0 |
Component: | StructureIdentification | Version: | 3.1 |
Keywords: | Cc: | michael.affenzeller@heuristiclab.com |
Description
Caching the return values of tree branches could improve GP evaluation performance. However this also involves a large memory overhead. Invalidation of caches of mutated tree-branches also adds complexity.
Careful consideration of positive and negative effects is necessary.
Change History (7)
comment:1 Changed 16 years ago by gkronber
- Cc michael.affenzeller@heuristiclab.com added
- Status changed from new to assigned
comment:2 follow-up: ↓ 3 Changed 16 years ago by gkronber
comment:3 in reply to: ↑ 2 Changed 16 years ago by gkronber
Replying to gkronber:
- Dataset: Wisconsin (683 samples)
- pop-size 700
- max-tree-size 100, max-tree-height 5
- Generations 10
- Evaluated Solutions 142600
- Evaluation: MSE
After some speed improvements:
- 5min 30sec (= 430 evals/sec)
comment:4 Changed 16 years ago by gkronber
- Resolution set to wontfix
- Status changed from assigned to closed
Experimented with caching the results of tree-branches on different levels (lowest level only vs. only branches with at least height 2). However the memory overhead was larger than the gain from cached evaluation.
comment:5 Changed 15 years ago by swagner
- Milestone changed from 3.1 to Iteration 0
Milestone 3.1 deleted
comment:6 Changed 13 years ago by swagner
- Milestone changed from Iteration 0 to Current
Milestone Iteration 0 deleted
comment:11 Changed 13 years ago by swagner
- Milestone changed from Current to HeuristicLab 3.3.0
Milestone Current deleted
Note: See
TracTickets for help on using
tickets.
first test run: