In Sagemaker catboost, what are the allowed parameters for learning rate?

Question

I am using Sagemaker out-of-the-box catboost algorithm and optimizing for learning rate. The official documentation states learning_rate hyperparameter can take values: MinValue: 0.001, MaxValue: 0.01.

This is clearly wrong -- for example one of the main authors advises a value of 0.03 is a good starting point. I tested values as high as 0.15 without throwing an error.

What is the actual maximum value or when can I find it? Thanks!

Answer 1

From detailed docs, it looks like maximum value is 1.0 (might be a typo in the page you shared) - https://docs.aws.amazon.com/sagemaker/latest/dg/catboost-hyperparameters.html