如何在Pytorch框架下使用tensorboardX:
[Pytorch]PyTorch使用tensorboardX（转
textCNN论文：
https://arxiv.org/abs/1408.5882
RCNN论文路径：
file:///Users/姓名/Downloads/9745-44425-1-PB.pdf
文本分类综述论文路径：
file:///Users/姓名/Downloads/information-10-00150-v2.pdf
多分类下的ROC曲线
指标介绍：AUC，ROC我看到的最透彻的讲解
ROC及AUC计算方法及原理
ROC原理介绍及利用python实现二分类和多分类的ROC曲线
https://stackoverflow.com/questions/39685740/calculate-sklearn-roc-auc-score-for-multi-class
scikit-learn官方文档-如何画多分类ROC曲线 一定要看这个！这个才是最好的！！！
https://www.jianshu.com/p/00ef5b63dfc8
碎碎念：
我觉得深度学习的本质就是维度。。。我又调了一晚上的维度，就画个ROC而已。。说明了我不适合写代码
在大佬的代码上补补改改，今日份的hello world:
import matplotlib.pyplot as plt
from sklearn.metrics import roc_curve, auc
from sklearn.preprocessing import label_binarize
from itertools import cycle
def evaluate(config, model, data_iter, test=False):
model.eval()
loss_total = 0
predict_all = np.array([], dtype=int)
labels_all = np.array([], dtype=int)
#######for roc#######
predict_all_roc = np.empty(shape=[0,10])
labels_all_roc = np.empty(shape=[0,10])
y_score_roc = np.empty(shape=[0,10])
with torch.no_grad():
for texts, labels in data_iter:
outputs = model(texts)
outputs_softmax = F.softmax(outputs,dim=1)
loss = F.cross_entropy(outputs, labels)
loss_total += loss
labels = labels.data.cpu().numpy()
predic = torch.max(outputs.data, 1)[1].cpu().numpy()
labels_all = np.append(labels_all, labels)
predict_all = np.append(predict_all, predic)
########for roc###########
labels_roc = label_binarize(labels, classes=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
labels_all_roc = np.concatenate((labels_all_roc, labels_roc), axis=0)
predic_roc = label_binarize(predic, classes=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
predict_all_roc = np.concatenate((predict_all_roc,predic_roc),axis=0)
y_score_roc = np.concatenate((y_score_roc,outputs_softmax),axis=0)
acc = metrics.accuracy_score(labels_all, predict_all)
if test:
# labels_all = label_binarize(labels_all,classes=[0,1,2,3,4,5,6,7,8,9])
fpr = {}
tpr = {}
roc_auc = {}
for i in range(10):
fpr[i], tpr[i], _ = roc_curve(labels_all_roc[:, i], y_score_roc[:,i])
roc_auc[i] = auc(fpr[i], tpr[i])
fpr['micro'],tpr['micro'],_ = roc_curve(labels_all_roc.ravel(),y_score_roc.ravel())
roc_auc['micro'] = auc(fpr['micro'],tpr['micro'])
lw = 2
plt.figure()
plt.plot(fpr[2], tpr[2],
label='micro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc[2]),
color='deeppink', linestyle='--', linewidth=4)
colors = cycle(['aqua', 'darkorange', 'cornflowerblue','azure','seashell','peru','mintcream','linen','indigo','gold'])
for i, color in zip(range(10), colors):
plt.plot(fpr[i], tpr[i], color=color, lw=lw,
label='ROC curve of class {0} (area = {1:0.2f})'
''.format(i, roc_auc[i]))
plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver operating characteristic example')
plt.legend(loc="lower right")
plt.show()
report = metrics.classification_report(labels_all, predict_all, target_names=config.class_list, digits=4)
confusion = metrics.confusion_matrix(labels_all, predict_all)
return acc, loss_total / len(data_iter), report, confusion
return acc, loss_total / len(data_iter)
# test(config=)
from importlib import import_module
from utils import build_dataset, build_iterator, get_time_dif
x = import_module('models.TextRNN_Att')
dataset = 'THUCNews'
embedding= 'random'
config = x.Config(dataset, embedding)
# print(vars(config))
vocab, train_data, dev_data, test_data = build_dataset(config,False)
#
test_iter = build_iterator(test_data, config)
model = x.Model(config).to(config.device)
# print(vars(model))
test(config,model,test_iter)
另外我真是黑人问号脸了，就这评价指标AUC能达到0.99，为了避免哪里问题我还把每一类都打印出来了，什么鬼啊，，这科学吗？？？？有没有大佬知道是什么情况啊。。科学不科学。。疯了
P-R-F1
ROC曲线
网友评论