Hi,
I am working on SVR. Input is current and output is resistance in step form. I tried a lot to train SVR manually and with PSO. when I use PSO, best cost never decreases.
Do I normalize input and output?
How can I return to original data when SVR is used in real-time system?
How can I find the best hyperparameters for SVR?
Is SVR an abbreviation of Support Vector Regression?
Me to I have used PSO to train SVR hyperparameters (you can check the published papers). But, I am not sure this could be always the best way to tune the hyperparameters. In many times, the available tools for instance in Scikit-learn could provide better results. This depends on the purpose of the paper.
I think in general normalization improve the performance of SVR. I don't know what do you mean by returning to original data, but normalization in general affects just the features and then the predictions remain the same.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training