I have a question about leave one out error obtained from PCE using uqlab:
I want to know how this error is calculated. I mean, considering we have N inputs and M number of evaluations of our function. Then we fit a PCE to our model and we get a leave one out error for that. How this error is calculated based on our inputs, and how we should analyze this number?
Thanks in advance
Thanks for your question.
First, it’s a good idea to read the UserManual_PCE carefully. There is a good discussion about the different types of errors.
Second, it’s highly recommended to read the fruitful answers of Prof. @bsudret to my question in this topic, Acceptable LOO Error to have a better insight about this magic tool.