# Surrogates -Polynomial Chaos Kriging

**URL:** <https://uqworld.org/t/surrogates-polynomial-chaos-kriging/588>\
**Category:** Community Q&A and How To\
**Created:** [May 8, 2020, 2:45pm UTC](https://uqworld.org/t/surrogates-polynomial-chaos-kriging/588 "2020-05-08T14:45:08Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![aokoro](https://uqworld.org/letter_avatar_proxy/v4/letter/a/49beb7/32.png) [@aokoro](https://uqworld.org/u/aokoro)\
**Post date:** [May 8, 2020, 2:45pm UTC](https://uqworld.org/t/surrogates-polynomial-chaos-kriging/588/1 "2020-05-08T14:45:08Z")

</div>

Are there any clear advantages of using Polynomial Chaos Kriging over surrogates like PCE, Kriging , SVR and ANN?

Appreciate your thoughts

---

<div class="post-metadata">

**Author:** ![GPLai](https://uqworld.org/user_avatar/uqworld.org/gplai/32/403_2.png) [@GPLai](https://uqworld.org/u/GPLai)\
**Post date:** [June 7, 2020, 3:54am UTC](https://uqworld.org/t/surrogates-polynomial-chaos-kriging/588/2 "2020-06-07T03:54:02Z")

</div>

Share some my pointview. There is a famous theory about the comparion between different Optimization Method, this theory named No Free Lunch Theory, which states that no one method outperform all others in every metric. This theory maybe applied for MetaModel area as well.

It would be hard to identify an optimum meta-model as abundance metrics may be of interest. And different people have very different preferences for these metrics. Someone may tend to select the simplest one, some people may prefer the most accurate one. I think one common rule maybe whether the method can solve our problem.

---

<div class="post-metadata">

**Author:** ![aokoro](https://uqworld.org/letter_avatar_proxy/v4/letter/a/49beb7/32.png) [@aokoro](https://uqworld.org/u/aokoro)\
**Post date:** [June 7, 2020, 11:48am UTC](https://uqworld.org/t/surrogates-polynomial-chaos-kriging/588/3 "2020-06-07T11:48:09Z")

</div>

Thank you so much Zhipeng for the insight.