shedtears4U / rattle

Automatically exported from code.google.com/p/rattle
0 stars 0 forks source link

Feature Request - Running All models to be enhanced - model scheduler idea ... #28

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
Hello, I do love the option to run All the models at once - especially as 
Rattle is pretty efficient using memory and CPU so I can run more models one 
after the other over night and I do not have to worry that much (not always 
though) that it will be interrupted by i.e. memory issues.

In this context I would like to suggest an idea of replacing/Enhancing the All 
radio button with a "model scheduler" which would allow to add particular model 
AND it's parameters to a queue which will be processed once you press the new 
"Run Scheduled Models" button.

This would allow you to not only to run several different models (rforrest, nn, 
glm, ...) automatically BUT ALSO allows to run 2 models of same category by one 
schedule (i.e. SVM with RBF Kernel and SVM with Polynomial kernel, or different 
sub-versions of linear models ... OR run 2 different Neural Networks trainings 
- one with X hidden units, 2nd with Y hidden units ... etc ...)

See some attached screenshots which are showing current limitations of ability 
to run different variants of same model categories.

Basic idea would be:

- Add button "Add to Model Scheduler" to store request to calculate particular 
model algorithm with it's all parameters needed for the model execution(model 
variant specific parameters)

- Add button "Run Scheduled Models" to start sequential execution of requested 
queue of models

- Optional - more work - have UI to maintain the queue of requested models (and 
their parameters) - then you can have the Run button with multiple selection 
options) here instead elsewhere

Original issue reported on code.google.com by cal...@gmail.com on 7 Jun 2014 at 2:29

Attachments: