参数优化-验证曲线

通过验证一个学习器在训练集和测试集上的表现，来确定模型是否合适，参数是否合适。

如果训练集和测试集得分都很低，说明学习器不合适。

如果训练集得分高，测试集得分低，模型过拟合，训练集得分低，测试集得分高，不太可能。

示例代码

import matplotlib.pyplot as plt
import numpy as np
from sklearn.datasets import load_digits
from sklearn.model_selection import validation_curve
from sklearn.svm import SVC

# 加载数据
digits = load_digits()
X, y = digits.data, digits.target

# 验证曲线
param_range = np.logspace(-6, -1, 5)
train_scores, test_scores = validation_curve(
    SVC(), X, y, param_name="gamma", param_range=param_range,
    cv=10, scoring="accuracy", n_jobs=1)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)

plt.title("SVM VC")
plt.xlabel("$gamma$")
plt.ylabel("Score")
plt.ylim(0.0, 1.1)

# 训练数据
plt.semilogx(param_range, train_scores_mean, label="train score", color="r")
plt.fill_between(param_range, train_scores_mean - train_scores_std,
                 train_scores_mean + train_scores_std, alpha=0.2, color="r")

# 测试数据
plt.semilogx(param_range, test_scores_mean, label="test score",color="g")
plt.fill_between(param_range, test_scores_mean - test_scores_std,
                 test_scores_mean + test_scores_std, alpha=0.2, color="g")
plt.legend(loc="best")
plt.show()

输出

参数gamma的调节

很小时，训练集和测试集得分都低，欠拟合

增大时，训练集和测试集得分有个很好地值

过大时，训练集得分高，测试集得分低，过拟合。