learn-etc

Describe your project here.

already compared get dummies lets see how sklearn onehot behaves on polars and pandas

flowchart TD
    a[GridSearchCV:: hyperparameter search] --> b[cv: cross-validation generator:: split strategy, e.g. time series split]

    aa[OptunaCV:: hyperparameter search] --> fit --> study --> optimize --> _objective 
    
    subgraph _objective
      cross_validate --> bb[cv: cross-validation generator:: split strategy, e.g. time series split] --> mean_score@{shape: doc, label: "mean_score"} --> self._store_scores

      self.cross_validate_with_pruning --> bb --> split_score@{shape : docs, label: "split_score"} --> pruner_related --> self._store_scores 
    end

  
    self._store_scores

    aaa[GridSearchCV] --> bbb[Pipeline] --> ccc[sampling ] --> classifier
    subgraph pruner_related
      trial.report
    end

Loading

optuna search:

• The goal is originally to modify the optuna_integration.sklearn.OptunaSearchCV to be feasible to multi objective searchs such that there is a sklearn compatible module.
• During research we also found this wrapper with similar intention: ray.tune.search.optuna.OptunaSearch
  • it support parametrized sampler, but maybe not pruner?
• ray tune can be integrated with mlflow to achieve experiment tracking.