For examples, how to choose the optimal number of latent variables and where to search for them? Where to search for SVMR hyperparameters? Considering five-fold leave-group-out cross-validation, can one tune separately in each fold of the hyperparameters, and if so, where will the result be validated?
the optimal number of latent variables can be fixed based on the cross validation error RMSECV as it should be as minimum as possible
The good performance of both methods in the standard SVMr with hyper-parameters and also in the extension to multi-parametric kernels approaches such as evolutionary algorithms must be used to look for the best set of SVMr hyper-parameters.
Comparisons with a standard K-fold cross validation as a reference method could be tested.
I think you need to do a search such as GridSearch for each of them and see how the model performance improve using the criteria of RMSE or R2.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps