TD_SVM produces two outputs for linear kernel and three outputs for non-linear radial basis function sampler kernel.:
TD_SVM produces two outputs:
- Model (primary output): Contains the trained model along with model statics. The model stores the following model statistics:
- Loss Function
- MSE
- Number of Observations
- AIC
- BIC
- Number of Iterations
- Regularization
- Alpha (L1/L2/Elasticnet)
- Learning Rate (Initial)
- Learning Rate (Final)
- Momentum
- Nesterov
- LocalSGD Iterations
- [Optional] MetaInformationTable (second output): Contains training progress information for each epoch.
- [Required if Kernel argument is 'RBF'] RBFFitTable (third output): Contains the weights and bias for features calculated.
Model Output Table Schema
| Column |
Data Type |
Description |
|---|
| attribute |
SMALLINT |
Contains a numeric index of predictor and model metrics. Intercept is specified using index 0 and rest of the predictors have positive values. Model metrics have negative indices. |
| predictor |
VARCHAR |
Contains the name of the predictor or model metric. |
| estimate |
FLOAT |
Contains the predictor weights and numeric-based metric values. |
| value |
VARCHAR |
Contains the string-based metric values for example, HINGE for LossFunction, L2 for Regularization, and so on. |
MetaInformationTable Output Table Schema [Optional]
| Column |
Data Type |
Description |
|---|
| iteration |
INTEGER |
Contains the iteration number (epoch number). |
| num_rows |
BIGINT |
Contains the total number of rows processed so far. |
| eta |
FLOAT |
Contains the learning rate for the iteration. |
| loss |
FLOAT |
Contains loss in this iteration. |
| best_loss |
FLOAT |
Contains best loss till this iteration. |
RBFFitTable Output Schema
| Column |
Data Type |
Description |
|---|
| Feature 1 |
FLOAT |
Contains weights for the feature. |
| ... |
... |
... |
| FeatureN |
FLOAT |
Contains weights for the feature. |
| Bias |
FLOAT |
Contains bias. |