ValueError: CategoricalDistribution does not support dynamic value space

Question

Created an objective function for optuna to find the best parameters for KNN regressor but faced with this error:

ValueError: CategoricalDistribution does not support dynamic value space

Any suggestions as to why this is happening?

def objective(trial):
    
    params = {
        'n_neighbors': trial.suggest_int('n_neighbors', 2, 10, step=2),
        'algorithm': trial.suggest_categorical('weights', ['auto', 'ball_tree', 'kd_tree', 'brute']),
        'weights': trial.suggest_categorical("weights", ['uniform', 'distance']),
        "leaf_size": trial.suggest_int("leaf_size", 10, 60, step=10),
        "p": trial.suggest_categorical("p", [1, 2]),
    }
    
    regression_model = KNeighborsRegressor(**params)
    regression_model.fit(x_train.values, y_train.values)
    
    y_pred = regression_model.predict(x_test)
    rmse = mean_squared_error(y_test, y_pred)
    
    return rmse

find_params = optuna.create_study(direction='minimize')
find_params.optimize(objective, n_trials=5)

Answer 1

You have "weights" parameter defined twice. Change from

    'algorithm': trial.suggest_categorical('weights', ['auto', 'ball_tree', 'kd_tree', 'brute']),
    'weights': trial.suggest_categorical("weights", ['uniform', 'distance']),

to

    'algorithm': trial.suggest_categorical('algorithm', ['auto', 'ball_tree', 'kd_tree', 'brute']),
    'weights': trial.suggest_categorical("weights", ['uniform', 'distance']),

and the error should go away. Optuna does not support having multiple parameters with same name but different value space.