Fix regression models

[Irozuku]

This pull request updates several regression model schemas and utility functions to improve parameter validation, consistency, and code clarity in both backend and frontend components. The main changes involve refining the types and constraints for model hyperparameters, updating metric computation to use the latest scikit-learn API, and enhancing frontend logic for handling optimizable hyperparameters.

Backend: Model Schema and Metric Updates

Regression model schema improvements:

• Changed min_weight_fraction_leaf and min_impurity_decrease in both RandomForestRegressionSchema and GradientBoostingRSchema to use float_field instead of optimizer_float_field, simplifying these fields and updating their placeholders accordingly. [1] [2] [3] [4]
• Adjusted constraints for several hyperparameters:
  • learning_rate now requires a minimum value of 0.01, and subsample a minimum of 0.1 in GradientBoostingRSchema. [1] [2]
  • C in LinearSVRSchema and alpha in RidgeRegressionSchema now use optimizer_int_field with a minimum of 1, and their placeholders were updated for integer values. [1] [2]
  • intercept_scaling in LinearSVRSchema now uses a minimum value of 1.0.
  • max_iter in RidgeRegressionSchema now requires a minimum of 10.

Regression metric update:

• Updated the RMSE metric implementation to use root_mean_squared_error from scikit-learn instead of computing RMSE via mean_squared_error(..., squared=False), aligning with the latest scikit-learn API. [1] [2]

Frontend: Hyperparameter Optimization Utilities

Hyperparameter optimizers logic:

• Added a new utility function checkHowManyOptimazers in schema.js to recursively count optimizable hyperparameters, including nested parameters.
• Updated ResultsTabHyperparameters.jsx to use the new checkHowManyOptimazers function for determining the number of optimizable hyperparameters, improving reliability for complex parameter schemas. [1] [2]