-
-
Notifications
You must be signed in to change notification settings - Fork 70
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improved GEKPLS Function #456
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #456 +/- ##
=======================================
Coverage 78.12% 78.12%
=======================================
Files 23 23
Lines 3154 3154
=======================================
Hits 2464 2464
Misses 690 690 ☔ View full report in Codecov by Sentry. |
@ChrisRackauckas |
|
||
# Modifications for Improved GEKPLS Function: | ||
|
||
To enhance the GEKPLS function, sampling method was changed from ```SobolSample()``` to ```HaltonSample()```. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's not an enhancement, that's just a change in the sampling method.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah but gekpls.jl might improve if we use Halton sampling instead of sobol sampling.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, but that's a user choice. That's not inherent to gekpls but just an option for it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes! Multiple choices might work for multiple applications. But at the core, we should focus on optimizing complex mathematical functions like matrix multiplication and also focus on optimal featurization which works better for real world applications.
Yes, but that's a user choice. That's not inherent to gekpls but just an option for it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What's your proof this is a better default? Can you show some head-to-head benchmarks?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I recently hit a PR, where I proposed changes in the readme file of tensor_prod.md
#457
But I think the kind of example functions that we are using are also not completely justified for the benchmark label.
We should use better examples to explain the surrogate optimization and as @sathvikbhagavan rightly said, we need to dive deeper into explaining the mathematical concepts.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, changes in tensor_prod.md
is in the right direction.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@sathvikbhagavan
So is it ready to merge or still do you think it can be changed at a few spots? Like I think I did not explain mathematically about the function and just textually explained the function.
Should I add in the whole concept image available online?
Link: https://projecteuclid.org/journals/tohoku-mathematical-journal/volume-17/issue-2/The-tensor-product-of-function-algebras/10.2748/tmj/1178243579.pdf
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can you clean up the PRs - use either this or #457 and close one of them, remove the extra tutorial which is not needed and then we can have a round of reviews?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@sathvikbhagavan and @ChrisRackauckas
Yeah so let's close this one as I am still experimenting with the sampling methods for better rmse!!
Will maybe again hit a PR for the same.
Update tensor_prod.md
Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
Add any other context about the problem here.