Opened 8 years ago
Closed 8 years ago
#2385 closed enhancement (done)
Boolean parameter for data-analysis algorithms to disable creation of a final solution
Reported by: | gkronber | Owned by: | gkronber |
---|---|---|---|
Priority: | medium | Milestone: | HeuristicLab 3.3.13 |
Component: | Algorithms.DataAnalysis | Version: | 3.3.11 |
Keywords: | Cc: |
Description
For multiple data-analysis algorithms the creation (+evaluation) of the final solution takes a lot of time.
Especially, when cross-validation is used this is however not necessary and only slows down CV. Therefore, it would be useful to add a Boolean (hidden) parameter to those algorithms to turn of creation of solutions.
The following algorithms would benefit:
- Random forest
- Gaussian process
- Support vector machine
Change History (11)
comment:1 Changed 8 years ago by ascheibe
- Owner set to gkronber
- Status changed from new to assigned
comment:2 Changed 8 years ago by gkronber
- Milestone changed from HeuristicLab 3.3.12 to HeuristicLab 3.3.13
comment:3 Changed 8 years ago by gkronber
comment:4 Changed 8 years ago by gkronber
- Status changed from assigned to accepted
comment:5 Changed 8 years ago by gkronber
r12934: added a Boolean "CreateSolution" parameter for support vector machine algorithms and added model error/accuracy metrics as algorithm results (to allow grid search without creating solutions)
comment:6 Changed 8 years ago by gkronber
r13204: added CreateSolution flag to random forest
comment:7 Changed 8 years ago by gkronber
r13205: added CreateSolution flag to Gaussian process algorithms
In the case of Gaussian process regression, it would be possible to exchange the solution creator of the problem. However, this cannot be easily achieved in the current version because of backwards-compatibility reasons. Therefore, I decided to use the same way of implementing this as for the other algorithms.
comment:8 Changed 8 years ago by gkronber
- Owner changed from gkronber to mkommend
- Status changed from accepted to reviewing
Please review the 3 changesets.
The flag has been added to
- SVM regression & classification
- RF regression & classification
- Gaussian process regression & classification
The flag was already available for gradient boosted trees for regression.
comment:9 Changed 8 years ago by mkommend
- Owner changed from mkommend to gkronber
- Status changed from reviewing to readytorelease
r13270: Removed wrong comment in SVM classification and regression.
comment:10 Changed 8 years ago by gkronber
comment:11 Changed 8 years ago by gkronber
- Resolution set to done
- Status changed from readytorelease to closed
Related: #2450, #2465