# Accuracy¶

class paddle.fluid.metrics.Accuracy(name=None)[源代码]

## 参数¶

• name (str，可选) – 具体用法请参见 Name ，一般无需设置，默认值为None。

Accuracy

## 代码示例¶

import paddle.fluid as fluid
# 假设有batch_size = 128
batch_size=128
accuracy_manager = fluid.metrics.Accuracy()
# 假设第一个batch的准确率为0.9
batch1_acc = 0.9
accuracy_manager.update(value = batch1_acc, weight = batch_size)
print("expect accuracy: %.2f, get accuracy: %.2f" % (batch1_acc, accuracy_manager.eval()))
# 假设第二个batch的准确率为0.8
batch2_acc = 0.8
accuracy_manager.update(value = batch2_acc, weight = batch_size)
#batch1和batch2的联合准确率为(batch1_acc * batch_size + batch2_acc * batch_size) / batch_size / 2
print("expect accuracy: %.2f, get accuracy: %.2f" % ((batch1_acc * batch_size + batch2_acc * batch_size) / batch_size / 2, accuracy_manager.eval()))
#重置accuracy_manager
accuracy_manager.reset()
#假设第三个batch的准确率为0.8
batch3_acc = 0.8
accuracy_manager.update(value = batch3_acc, weight = batch_size)
print("expect accuracy: %.2f, get accuracy: %.2f" % (batch3_acc, accuracy_manager.eval()))


## 方法¶

### update(value, weight)¶

$\begin{split}\\ \begin{array}{l}{\text { self. value }+=\text { value } * \text { weight }} \\ {\text { self. weight }+=\text { weight }}\end{array} \\\end{split}$

• value (float|numpy.array) – mini-batch的正确率
• weight (int|float) – mini-batch的大小