KNN算法识别手写数字识别集

调用sklearn中KNN算法库，图片数据级为matlab文件，使用matplot读取，并进行一维化处理

from scipy.io import loadmat as load
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier as kNN
import numpy as np


# 加载matlab格式数据文件
train = load('train_32x32.mat')
test = load('test_32x32.mat')

train_samples = train['X']
train_labels = train['y']

test_samples = test['X']
test_labels = test['y']


def reformat(samples, labels):
    """
    改变原始数据的形状
    (图片高，图片宽，通道数，图片数）-> （图片数，图片高，图片宽，通道数）
    把labers转化为列向量
    """
    # 改变张量形状信息
    samples = np.transpose(samples, (3, 0, 1, 2))


  
    

    # 改变标签信息
    labels = np.array([x[0] for x in labels])
    one_hot_labels = []
    for num in labels:
        one_hot =[0.0] * 10
        if num == 10:
            one_hot[0] = 1.0
        else:
            one_hot[num] = 1.0
        one_hot_labels.append(one_hot)
    labels = np.array(one_hot_labels).astype(np.float32)
    
    
    

    return samples, labels




def normalize(samples):
    """
    将图片灰度化：三色通道到单通道
    将图片 0到255线性映射到 -1.0到1.0
    """
    samples = np.add.reduce(samples, keepdims = True, axis = 3)
    samples = samples/128 - 1
    return samples


def inspect(data_set, labels, i):
    """显示第i张数据集图片"""
    print(labels[i])
    plt.imshow(data_set[i])
    plt.show()


if __name__ == '__main__':
    train_samples,train_labels= reformat(train_samples, train_labels)
    train_samples = normalize(train_samples)

    # 一维化图像
    train_samples = train_samples.reshape(train_samples.shape[0], -1)


    test_samples, test_labels = reformat(test_samples, test_labels)
    test_samples = normalize(test_samples)


    # 一维化图像
    test_samples = test_samples.reshape(test_samples.shape[0], -1)

    # 取100张进行预测
    test = test_samples[0:1000, :]
    y = test_labels[0:1000, :]
    knn_clf = kNN(n_neighbors=5)
    print(knn_clf.fit(train_samples, train_labels))
    y_ = knn_clf.predict(test)

    rate = (y_ == y).sum()/y_.size
    print("正确率为%f"%rate)