大作业，细节以后再补

USPS

import h5py
import numpy as np
import cv2
import torch

# load data
# data from https://www.kaggle.com/bistaumanga/usps-dataset?select=usps.h5
# another data source https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass.html#usps
path = 'usps.h5'
with h5py.File(path, 'r') as hf:
    train = hf.get('train')
    x_tr = train.get('data')[:]
    y_tr = train.get('target')[:]
    test = hf.get('test')
    x_te = test.get('data')[:]
    y_te = test.get('target')[:]


def check_pic(data, label, idx):
    pic = data[idx, :]
    pic = pic.reshape((16, 16))
    cv2.imshow(str(label[idx]), pic)
    cv2.waitKey(0)


# check picture
# check_pic(x_tr, y_tr, 10)


# KNN
def knn(x, k):
    idx = np.argsort(np.sum((x_tr - x_te[x, :]) ** 2, axis=1))
    vote = np.zeros(10)
    for i in range(k):
        vote[y_tr[idx[i]]] += 1
    return np.argmax(vote)


for k in range(1, 20):
    err = 0
    for x in range(x_te.shape[0]):
        y_pr = knn(x, k)
        err += y_pr != y_te[x]
        print(x)
    print(k, 1 - err / x_te.shape[0])

USI sonar

import numpy as np
import pandas as pd

# load data
# data from http://archive.ics.uci.edu/ml/machine-learning-databases/undocumented/connectionist-bench/sonar/
path = 'sonar.all-data'
data = pd.read_csv(path).values
labels = data[:, -1]
data = data[:, :-1]

train_rate = 0.7
idx = np.arange(labels.shape[0])
idx = [np.random.permutation(idx[labels == 'R']), np.random.permutation(idx[labels == 'M'])]
train_idx = [idx[0][:int(train_rate * idx[0].shape[0])], idx[1][:int(train_rate * idx[1].shape[0])]]
train_idx = np.concatenate((train_idx[0], train_idx[1]))
test_idx = [idx[0][int(train_rate * idx[0].shape[0]):], idx[1][int(train_rate * idx[1].shape[0]):]]
test_idx = np.concatenate((test_idx[0], test_idx[1]))
x_tr = data[train_idx, :]
y_tr = labels[train_idx]
x_te = data[test_idx, :]
y_te = labels[test_idx]


# KNN
def knn(x, k):
    idx = np.argsort(np.sum((x_tr - x_te[x, :]) ** 2, axis=1))
    vote = {'M': 0, 'R': 0}
    for i in range(k):
        vote[y_tr[idx[i]]] += 1
    return 'M' if vote['M'] > vote['R'] else 'R'


for k in range(1, 20):
    err = 0
    for x in range(x_te.shape[0]):
        y_pr = knn(x, k)
        err += y_pr != y_te[x]
    print(k, 1 - err / x_te.shape[0])

USI iris

from sklearn import datasets
import numpy as np

iris = datasets.load_iris()
data = iris['data']
labels = iris['target']

train_rate = 0.7
idxx = np.arange(labels.shape[0])
train_idx = []
test_idx = []
for i in range(3):
    idx = np.random.permutation(idxx[labels == i])
    train_idx.append(idx[:int(train_rate * idx.shape[0])])
    test_idx.append(idx[int(train_rate * idx.shape[0]):])
train_idx = np.concatenate(tuple(train_idx))
test_idx = np.concatenate(tuple(test_idx))
x_tr = data[train_idx, :]
y_tr = labels[train_idx]
x_te = data[test_idx, :]
y_te = labels[test_idx]


# KNN
def knn(x, k):
    idx = np.argsort(np.sum((x_tr - x_te[x, :]) ** 2, axis=1))
    vote = np.zeros(3)
    for i in range(k):
        vote[y_tr[idx[i]]] += 1
    return np.argmax(vote)


for k in range(1, 20):
    err = 0
    for x in range(x_te.shape[0]):
        y_pr = knn(x, k)
        err += y_pr != y_te[x]
    print(k, 1 - err / x_te.shape[0])