Cookbook for Practice 3.1

Cong Li 李聪

实践3.1攻略

Nearest Neighbor 最近邻 (1)

Start a new file ‘nearestNeighbor.py’ w/ a new class. Follow the training part in rote learning.

创建一个包含新的类型的新文件“nearestNeighbor.py”。训练部分参照死记硬背的学习。

Nearest Neighbor 最近邻 (2)

Now we write the ‘classify’ function, starting w/ a local variable for minimal distance & another for the nearest data sample seen so far

现在我们写“classify”函数，为到目前为止遇到的最小距离和最邻近的数据各准备一个局部变量

Nearest Neighbor 最近邻 (3)

Loop over all the training data samples, calculate the distance, & update the minimum distance & the nearest neighbor on demand. Return the class label of the nearest neighbor at the end.

遍历所有的训练数据，计算距离，并根据情况更新最小距离和最近邻。最后返回最近邻的类别。

Run It 运行之

In ‘tester.py’, print something for every 100 data samples classified 在“tester.py”中，每分类100个数据就打印一下

In ‘test.py’, import ‘nearestNeighbor’, follow rote learning to create an instance for performance evaluation

在“test.py”中，导入“nearestNeighbor”，参照死记硬背的学习，创建其中类型的实例，对其分类性能进行评估

Train on ‘usps’, test on ‘usps.t’ (& be very patient)

用“usps”进行训练，用“usps.t”进行评估（并表现出异常的耐心）

The End