Some issues related to sgp_chs.ipynb code
Closed this issue · 1 comments
Hello author, I am glad that you have done an excellent job of open-source. I am a beginner and after watching your sgp_chs.ipynb GP tutorial, I believe that there are two issues with the code in this tutorial, which leads to inconsistencies with the mathematical formulas you provided in the tutorial.
1.When constructing the kernel function and model parameters, the following code has already multiplied the a scaling factor, but when it returns, the K matrix is multiplied by a scaling factor a. Is there a discrepancy with the formula here?
2.In another better method of writing kernel functions, during the process of data standardization, you divide x by the square of ld before multiplying x. This process results in the denominator in the final formula being divided by the fourth power of ld, while the formula shows dividing by the square of ld. Is there a problem here?
Looking forward to your reply!
It has been modified in zhenjie's branch