Estimator file change #72
Unanswered
olivierlabayle
asked this question in
Q&A
Replies: 3 comments 1 reply
-
Hi Olivier, this is fine with me! I think that it would be useful to maybe link to the documentation of these functions or provide an example of each maybe in the documentation for the targene-pipeline just so that people can easily adapt them to their liking for future use. |
Beta Was this translation helpful? Give feedback.
0 replies
-
Hey,
I guess my main questions is whether this would change anything for the user apart from having to write Julia files rather than yaml before a run?
On 13 Jun 2023, at 09:35, Olivier Labayle ***@***.***> wrote:
This email was sent to you by someone outside the University.
You should only click on links or attachments if you are certain that the email is genuine and the content is safe.
Hi @TARGENE/contributors<https://github.com/orgs/TARGENE/teams/contributors>,
I am thinking of switching the estimator configuration file which is currently a .yaml file to plain julia file (.jl). The idea is to enable more models out of the box without having to write an explicit way to convert between yaml and julia in the code. This would notably help to natively support all composite models in MLJ<https://alan-turing-institute.github.io/MLJ.jl/dev/> like Pipelines and TunedModels.... It would also make the syntax probably simpler in general. As an example, the following specification for the Q learner:
Q_continuous:
model: Stack
# Description of the resampling strategy
resampling:
type: CV
nfolds: 2
# List all models and hyperparameters
models:
- type: GLMNetRegressor
order: 3
- type: EvoTreeRegressor
nrounds: 10
- type: LinearRegressor
- type: HALRegressor
would become:
Q_continuous = Stack(
resampling=CV(nfolds=2),
metalearner=LinearRegressor(),
ev=EvoTreeRegressor(nrounds=10),
lr=LinearRegressor(),
hal=HALRegressor(),
glmnet=GLMNetRegressor)
Before I do anything I wanted to make sure this change is fine with you?
—
Reply to this email directly, view it on GitHub<#72>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AXIXLMAFX2ASPDHAMYMVPGDXLAQ4XANCNFSM6AAAAAAZEP5MMI>.
You are receiving this because you are on a team that was mentioned.Message ID: ***@***.***>
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. Is e buidheann carthannais a th’ ann an Oilthigh Dhùn Èideann, clàraichte an Alba, àireamh clàraidh SC005336.
|
Beta Was this translation helpful? Give feedback.
1 reply
-
Note that I will update the current "classic" estimator files here that we have in TarGene to fit that new design. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi @TARGENE/contributors,
I am thinking of switching the estimator configuration file which is currently a .yaml file to plain julia file (.jl). The idea is to enable more models out of the box without having to write an explicit way to convert between yaml and julia in the code. This would notably help to natively support all composite models in MLJ like Pipelines and TunedModels.... It would also make the syntax probably simpler in general. As an example, the following specification for the Q learner:
would become:
Before I do anything I wanted to make sure this change is fine with you? You can simply react to this in order to vote or provide a more detailed answer if any :)
Beta Was this translation helpful? Give feedback.
All reactions