[Bug] Problems in the normalization and standardization of data
JoseAero opened this issue · 1 comments
🐛 Bug
Hello,
I am trying to compare the results of a simple model that predicts the heat of a fluid by taking into account its dimensionless velocity, dimensionless local temperature, dimensionless density, dimensionless pressure and dimensionless viscosity (5 features). I want to compare the results in the case of normalizing the features and standardizing the label and the case of not applying these transformations.
In principle I want to do it with noise fixed since the error of the computational code from which the data have been taken can be 5% of the value of the corresponding label.
The problem is that when I compare the data, the standardized case does not seem to work and I do not understand why.
I would appreciate some light on this.
Attached is the code with the databases.
Thanks in advance and best regards
System information
Please complete the following information:
- GPyTorch Version: 1.11
- PyTorch Version: 2.0.1
- Window 11
Normalizing the test seems to work better but with greater uncertainty, and the 5-element training case continues without improvement