Spinachboul update function #457

Spinachboul · 2023-12-28T11:37:42Z

Changes Made

Explained Tensor Product Functions for the user's reference.
Changed the function into a more readable form.
Attached the relevant plots for different Surrogate Models for visualization.

codecov · 2023-12-28T11:50:17Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (aa744e5) 78.12% compared to head (688fee1) 78.12%.
Report is 13 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #457   +/-   ##
=======================================
  Coverage   78.12%   78.12%           
=======================================
  Files          23       23           
  Lines        3154     3155    +1     
=======================================
+ Hits         2464     2465    +1     
  Misses        690      690

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Spinachboul · 2023-12-30T14:39:49Z

@ChrisRackauckas and @sathvikbhagavan
Please have a look at the changes made in the readme file and suggest anything which could be improved!!

ChrisRackauckas · 2023-12-30T16:54:48Z

docs/src/tensor_prod.md

+![kriging](https://github.com/Spinachboul/Surrogates.jl/assets/105979087/906e6688-db47-48be-90d1-ea471aacac16)
+
+## Lobachevsky Plot
+
+![lobachevsky](https://github.com/Spinachboul/Surrogates.jl/assets/105979087/678cfc13-0aec-4488-8e4d-39649853ecdd)
+
+## Combined Plot
+
+![combined_plot](https://github.com/Spinachboul/Surrogates.jl/assets/105979087/46762f0d-50c5-4d6c-961a-236fd9fb3ad5)


This is unrelated?

Oh I see, but the docs already generate these plots.

Yeah, that's right! I attached the plots for reference about Kriging and Lobachevsky individually, and their plotting curves. And can't we attach plots both, in the readme and the docs??

No, I don't recommend attaching plots but instead running code to generate them

ChrisRackauckas · 2023-12-30T16:55:19Z

docs/src/Improvedgekpls.md

+# GEKPLS Function
+
+Gradient Enhanced Kriging with Partial Least Squares Method (GEKPLS) is a surrogate modelling technique that brings down computation time and returns improved accuracy for high-dimensional problems. The Julia implementation of GEKPLS is adapted from the Python version by [SMT](https://github.com/SMTorg) which is based on this [paper](https://arxiv.org/pdf/1708.02663.pdf).  
+
+# Modifications for Improved GEKPLS Function:
+
+To enhance the GEKPLS function, sampling method was changed from ```SobolSample()``` to ```HaltonSample()```.
+
+
+```@example gekpls_water_flow
+
+using Surrogates
+using Zygote
+
+function water_flow(x)


This is just GEKPLS, why would it have a different page?

Yeah I added that in the Pages.jl for the modifications I made in the new file, I will remove it!

ChrisRackauckas · 2023-12-30T16:55:35Z

docs/src/Improvedgekpls.md

+| **Sampling Method** | **RMSE**                | **Differences**                                                                                                                                                                                                                                                                                     |
+|----------------------|--------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| **Sobol Sampling**               | 0.021472963465423097 | Utilizes digital nets to generate quasi-random numbers, offering low discrepancy points for improved coverage. - Requires careful handling, especially in higher dimensions.                                                                                                                      |
+| **Halton Sampling**               | 0.02144270998045834  | Uses a deterministic sequence based on prime numbers to generate points, allowing for quasi-random, low-discrepancy sampling. - Simpler to implement but may exhibit correlations in some dimensions affecting coverage.


This is more for QuasiMonteCarlo.jl

sathvikbhagavan · 2023-12-31T04:56:19Z

docs/src/Improvedgekpls.md

Like we talked about in #456, GEKPLS is independent from sampling method and the notion of improvement in accuracy depends on the original problem itself and what sampling works better. There is no general recipe for making it better. So, I don't think this tutorial is adding any new value and should be removed.

@sathvikbhagavan
Your'e right! If anything, we could have better optimization techniques to improve in gekpls.jl, instead of solely focusing on sampling methods! So, I have removed the file.

I don't think you have removed it. You might have forgot to push it. Look at the diff and commits to confirm.

delete Improvedgekpls.md and then we can merge this PR.

Yeah, I have deleted the file Improvedgekpls.md
Please check once again if all the changes are suitable for merging.

@ChrisRackauckas and @sathvikbhagavan
Is this PR ready for merging? Or still there are things which need to be modified??

sathvikbhagavan · 2023-12-31T04:57:40Z

docs/src/tensor_prod.md

+
+## Kriging Plot
+
+![kriging](https://github.com/Spinachboul/Surrogates.jl/assets/105979087/906e6688-db47-48be-90d1-ea471aacac16)


I don't think this is necessary. The plots would be generated from the above code block. Are you doing anything new, as in what code produces this plot?

I didn't do anything new logically in the code, just added in some explanation about the tensor product function which I thought would be relevant and added the plots , i will remove if it is not required!!

Yes, remove them

@sathvikbhagavan
But then, shouldn't we add more mathematical explanation to improve our readme file, simply stating the function would be too vague for the viewers!

I don't understand the usefulness of the plots. They don't provide any new information either. Look at the current plot in the tutorial - https://docs.sciml.ai/Surrogates/stable/tensor_prod/

Ok, so for now, I just need to focus on adding mathematical explanations in the tutorials right?
So soon I'll share a sample file (won't hit a PR for this) and then request your approval if it is going along the correct lines.
Please then suggest to me the correct blend of theory and math explanations which would enable me to work efficiently.

Sure. Feel free to open up a PR and continue working. Make it a draft PR and you can push in commits. When you are ready, make it ready for review and ping me. I will take a look. Hopefully you are more comfortable in the ecosystem and how things work in SciML.

Yeah! I am actually getting to know SciML codebase better. Thanks for your guidance throughout this time.

Hey @ChrisRackauckas and @sathvikbhagavan
Please let me know if this explanation suits well with the Kriging Model!

I was wondering if this is how we could expand the docs in the tutorials section!!
(Please forgive for the bad handwriting!! 😅)

And please also suggest models on which I could work! I mean there are many out there but the ones which are important and really catch reader's eye.

docs/src/tensor_prod.md

sathvikbhagavan · 2023-12-31T05:10:13Z

docs/src/tensor_prod.md

+
+For instance, consider a tensor product function defined as follows:
+
+```\[ f(x) = ∏ᵢ=₁ᵈ cos(aπxᵢ) \]```


Is this different from what it was before?

No, it is the same just in a more readable form than before, the limits and the function, which we usually write!
As earlier it was a bit confusing as to what the actual function is!

Alright, then its fine

Co-authored-by: Sathvik Bhagavan <[email protected]>

ChrisRackauckas · 2024-01-04T13:05:44Z

docs/src/tensor_prod.md

@@ -1,6 +1,5 @@
 # Tensor product function
-The tensor product function is defined as:
-``f(x) = \prod_{i=1}^d \cos(a\pi x_i)``
+The tensor product function is defined as: ```\[ f(x) = ∏ᵢ=₁ᵈ cos(aπxᵢ) \]```


But this isn't correct. This is one example of a tensor product.

Yes, that is true. Maybe we specifically say it is one of the possible forms. I think this was taken from https://smt.readthedocs.io/en/latest/_src_docs/problems/tensorproduct.html

Well, @ChrisRackauckas and @sathvikbhagavan
Should we properly define the tensor product function like this:

Or just stating this example would be fine??

Yes that would do it. But do we have an actual issue that is being solved here?

As the benchmarks need some cleanup, I think we can startoff with this PR to clean up the tensor product benchmark? Although I am not sure why would we need a surrogate for computing tensor product 😅

@sathvikbhagavan
I think there would be a couple of factors as to why surrogates could be used to compute tensor product:

Computational Complexity would be decreased

Efficient memory handling

Normalization of products of tensors which provide numerical stability

This PR #457 only consists of changing the example of tensor product into a more readable form up till now, but I would also work towards explaining the mathematical concepts behind surrogate models in my draft PR as we discussed earlier.

Spinachboul and others added 5 commits December 21, 2023 14:17

Improved GEKPLS Function

07c0131

Update Improvedgekpls.md

4358bee

Update Improvedgekpls.md

ef4c12b

Update pages.jl

da68799

Update tensor_prod.md

2e519c9

Spinachboul mentioned this pull request Dec 29, 2023

Improved GEKPLS Function #456

Closed

5 tasks

ChrisRackauckas reviewed Dec 30, 2023

View reviewed changes

sathvikbhagavan reviewed Dec 31, 2023

View reviewed changes

Spinachboul and others added 6 commits December 31, 2023 15:40

Update docs/src/tensor_prod.md

c546840

Co-authored-by: Sathvik Bhagavan <[email protected]>

Update pages.jl

dc946d4

Update tensor_prod.md

63a72d3

Update tensor_prod.md

670ecc7

Update Improvedgekpls.md

02aabfb

Delete docs/src/Improvedgekpls.md

fa8164a

ChrisRackauckas reviewed Jan 4, 2024

View reviewed changes

Update tensor_prod.md

688fee1

Spinachboul closed this by deleting the head repository Jan 7, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Spinachboul update function #457

Spinachboul update function #457

Spinachboul commented Dec 28, 2023

codecov bot commented Dec 28, 2023 •

edited

Loading

Spinachboul commented Dec 30, 2023

ChrisRackauckas Dec 30, 2023

ChrisRackauckas Dec 30, 2023

Spinachboul Dec 31, 2023

sathvikbhagavan Dec 31, 2023

ChrisRackauckas Dec 30, 2023

Spinachboul Dec 31, 2023

ChrisRackauckas Dec 30, 2023

sathvikbhagavan Dec 31, 2023

Spinachboul Dec 31, 2023

sathvikbhagavan Dec 31, 2023

sathvikbhagavan Jan 4, 2024

Spinachboul Jan 4, 2024

Spinachboul Jan 7, 2024

sathvikbhagavan Dec 31, 2023

Spinachboul Dec 31, 2023

sathvikbhagavan Dec 31, 2023

Spinachboul Dec 31, 2023

sathvikbhagavan Dec 31, 2023

Spinachboul Jan 4, 2024

sathvikbhagavan Jan 4, 2024

Spinachboul Jan 4, 2024

Spinachboul Jan 5, 2024 •

edited

Loading

Spinachboul Jan 6, 2024

sathvikbhagavan Dec 31, 2023

Spinachboul Dec 31, 2023

sathvikbhagavan Dec 31, 2023

ChrisRackauckas Jan 4, 2024

sathvikbhagavan Jan 4, 2024

Spinachboul Jan 4, 2024 •

edited

Loading

ChrisRackauckas Jan 4, 2024

sathvikbhagavan Jan 5, 2024

Spinachboul Jan 5, 2024


		## Kriging Plot

		![kriging](https://github.com/Spinachboul/Surrogates.jl/assets/105979087/906e6688-db47-48be-90d1-ea471aacac16)


		For instance, consider a tensor product function defined as follows:

		```\[ f(x) = ∏ᵢ=₁ᵈ cos(aπxᵢ) \]```

Spinachboul update function #457

Spinachboul update function #457

Conversation

Spinachboul commented Dec 28, 2023

Changes Made

codecov bot commented Dec 28, 2023 • edited Loading

Codecov Report

Spinachboul commented Dec 30, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Spinachboul Jan 5, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Spinachboul Jan 4, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov bot commented Dec 28, 2023 •

edited

Loading

Spinachboul Jan 5, 2024 •

edited

Loading

Spinachboul Jan 4, 2024 •

edited

Loading