custom kernel for GPR

2 ビュー (過去 30 日間)
Urmila Datta
Urmila Datta 2022 年 2 月 18 日
回答済み: Kausthub 2024 年 2 月 1 日
I have written custom kernel for GP regression with sum of squaredexponential+whitenoise and a constant.
Working perfect when predicting the same input after gpfit. But when I upsampled the input,it is showing Arrays have incompatible sizes for this operation. But when I correct the matrix size by breaking with dbstop error, it shows error in predict and showing Error using * . Attaching my custom kernel fuction.

回答 (1 件)

Kausthub
Kausthub 2024 年 2 月 1 日
Hi Urmila,
I believe that you are facing an error while upsampling the input and passing it to the custom kernel for GPR. Since, you have upsampled the input, the dimensions of the input changes and leads to the "Arrays have incompatibe sizes for this operation" error. From the error message I believe that it looks like a case of dimension mismatch for the matrix multiplication operation ‘*’. I would suggest you to:
1) Cross check the dimensions of your arrays and ensure that their dimensions are compatible for the matrix multiplication operation. Below mentioned are some documentations that might be useful for you apart from using breakpoints to solve the issue:
2) Cross check whether you actually wanted to do an element wise multiplication instead of a matrix multiplication. Here is a reference to a MATLAB Answer with the exact error message which might be helpful:
Hope it helps!

カテゴリ

Help Center および File ExchangeGaussian Process Regression についてさらに検索

タグ

製品


リリース

R2021b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by