Question

我正在尝试与python中的joblib库并行执行交叉验证折叠。

我有以下示例代码：

from sklearn.model_selection import KFold
import numpy as np
from sklearn.metrics import classification_report, confusion_matrix, f1_score
from sklearn import svm
from sklearn import datasets
from sklearn.model_selection import StratifiedKFold
from sklearn.svm import LinearSVC

iris = datasets.load_iris()
X, Y = iris.data, iris.target

skf = StratifiedKFold(n_splits=5)

#clf = svm.LinearSVC()
clf = svm.SVC(kernel='rbf')
#clf = svm.SVC(kernel='linear')

f1_list = []

for train_index, test_index in skf.split(X, Y):
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = Y[train_index], Y[test_index]
    clf.fit(X_train, y_train)
    Y_predict = clf.predict(X_test)
    f1 = f1_score(y_test, Y_predict, average='weighted')
    print(f1)
    conf_mat = confusion_matrix(y_test, Y_predict)
    print(conf_mat)
    f1_list.append(f1)
print(f1_list)

我想并行执行for循环，以并行获取每个折叠的准确性得分。

我认为joblib库必须通过以下方式使用：

from math import sqrt
from joblib import Parallel, delayed
def producer():
     for i in range(6):
        print('Produced %s' % i)
        yield i
out = Parallel(n_jobs=2, verbose=100, pre_dispatch='1.5*n_jobs')(
                delayed(sqrt)(i) for i in producer())

关于如何完成并行任务集成的任何建议？

Answer 1

在Parallel构造函数中，使用delayed参数指定要并行运行的函数。 delayed返回包装您的函数的新函数。然后，您可以使用将传递给原始函数的参数来调用新包装的函数。

在您的示例中，sqrt函数由delayed包装，然后从i并行发送range(6)。

我们需要做的是传递delayed一个可以训练大量数据的函数，然后将该新包装的函数传递给kfold拆分的索引。这是执行此操作的示例：

from sklearn.model_selection import KFold
import numpy as np
from sklearn.metrics import classification_report, confusion_matrix, f1_score
from sklearn import svm
from sklearn import datasets
from sklearn.model_selection import StratifiedKFold
from sklearn.svm import LinearSVC
from joblib import Parallel, delayed

iris = datasets.load_iris()
X, Y = iris.data, iris.target

skf = StratifiedKFold(n_splits=5)
clf = svm.SVC(kernel='rbf')

def train(train_index, test_index):
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = Y[train_index], Y[test_index]

    clf.fit(X_train, y_train)
    Y_predict = clf.predict(X_test)

    f1 = f1_score(y_test, Y_predict, average='weighted')
    conf_mat = confusion_matrix(y_test, Y_predict)

    return dict(f1=f1, conf_mat=conf_mat)


out = Parallel(n_jobs=2, verbose=100, pre_dispatch='1.5*n_jobs')(
    delayed(train)(train_index, test_index) for train_index, test_index in skf.split(X, Y))

f1_scores = [d['f1'] for d in out]
conf_mats = [d['conf_mat'] for d in out]

print('f1_scores:', f1_scores)
print('confusion matrices:', conf_mats)

出局：

f1_scores: [0.9665831244778613, 1.0, 0.9665831244778613, 0.9665831244778613, 1.0]
confusion matrices: [array([[10,  0,  0],
       [ 0, 10,  0],
       [ 0,  1,  9]], dtype=int64), array([[10,  0,  0],
       [ 0, 10,  0],
       [ 0,  0, 10]], dtype=int64), array([[10,  0,  0],
       [ 0,  9,  1],
       [ 0,  0, 10]], dtype=int64), array([[10,  0,  0],
       [ 0,  9,  1],
       [ 0,  0, 10]], dtype=int64), array([[10,  0,  0],
       [ 0, 10,  0],
       [ 0,  0, 10]], dtype=int64)]

out包含train函数返回的指标，因此我们可以根据需要将f1分数和混淆矩阵分开。

如何将joblib与scikitlearn并行使用交叉验证

1 个答案: