Hello, I am training a regression model with ridge regression and self-defined samples. The same heldout data are used twice, so each model is outputting the predictions for each data twice. I would expect that the predictions obtained for the same heldout data to be the same, yet they are slightly different.
Below you have a screenshoot, where the same rowIndex refer to the same heldout data, and the corresponding predictions.
In the manual, I found the following:
"For particular model, a grid of parameters (if any) is created and the model is trained on slightly different data for each candidate combination of tuning parameters. "
I suspect that the observed slightly different predictions may be related to this. My question here is do you modify the data before training/prediction? If so, what is the purpose?
Hello, I am training a regression model with ridge regression and self-defined samples. The same heldout data are used twice, so each model is outputting the predictions for each data twice. I would expect that the predictions obtained for the same heldout data to be the same, yet they are slightly different.
Below you have a screenshoot, where the same rowIndex refer to the same heldout data, and the corresponding predictions.
In the manual, I found the following:
I suspect that the observed slightly different predictions may be related to this. My question here is do you modify the data before training/prediction? If so, what is the purpose?