套索超参数整定研究中的最优收敛警告

Question

当使用Optuna微调我的套索模型时，我得到以下ConvergenceWarning。。是否有可能增加迭代次数？我增加了n_试验，但没有帮助

我的代码：

def objective(trial):

    _alpha = trial.suggest_float("alpha", 0.0001, 0.01)
    lasso = Lasso(alpha=_alpha, random_state=random_state)
    score = cross_val_score(lasso, X_train, y_train, cv=kf, scoring="neg_root_mean_squared_error").mean()

    return score


optuna.logging.set_verbosity(0)

study = optuna.create_study(direction="maximize")
study.optimize(objective, n_trials=300)

错误/警告：

/opt/conda/lib/python3.7/site-packages/sklearn/linear_model/_coordinate_descent.py:532: ConvergenceWarning:

Objective did not converge. You might want to increase the number of iterations. Duality gap: 0.02194362081235468, tolerance: 0.01627441311545211

Answer 1

根据警告消息，我认为代码中的Lasso是sklearn的^{}。因此，optuna的n_trials无法解决警告问题

我认为指定大于默认值的^{}的max_iter或tol参数可以解决警告问题