Codes for Machine Learning Foundations(NTU) 台湾国立大学《机器学习基石》（Coursera版）相关的代码、编程作业等。

hw1_naive_perceptron.py

#!/usr/bin/env python
#coding=utf8

""" DATA: https://d396qusza40orc.cloudfront.net/ntumlone%2Fhw1%2Fhw1_15_train.dat
Each line of the data set contains one (xn,yn) with xn∈R4. T
he first 4 numbers of the line contains the components of xn orderly, the last number is yn.
Please initialize your algorithm with w=0 and take sign(0) as −1

Question 15:
    Implement a version of PLA by visiting examples in the naive cycle using the
  order of examples in the data set. Run the algorithm on the data set.
  What is the number of updates before the algorithm halts?

Question 16:
    Implement a version of PLA by visiting examples in fixed, pre-determined random
  cycles throughout the algorithm. Run the algorithm on the data set. Please repeat
  your experiment for 2000 times, each with a different random seed. What is the average
  number of updates before the algorithm halts?

Question 17:
    Implement a version of PLA by visiting examples in fixed, pre-determined random cycles
  throughout the algorithm, while changing the update rule to be:
        wt+1 = wt + alpha * yn(t)xn(t)
  with alpha =0.5. Note that your PLA in the previous Question corresponds to alpha=1.
  Please repeat your experiment for 2000 times, each with a different random seed.
  What is the average number of updates before the algorithm halts?
"""

import random
from numpy import array, inner, zeros

DATA_FILE = 'ntumlone_hw1_hw1_15_train.dat'


def sign(x):
    if x <= 0:
        return -1
    return 1


def load_data(infile):
    X = []
    Y = []
    with open(infile) as f:
        for line in f:
            recs = line.split()
            x = [1] + [float(v) for v in recs[:-1]]
            X.append(tuple(x))
            Y.append(int(recs[-1]))
    return array(X), array(Y)


def train(X, Y, rand=False, alpha=1):
    n = len(Y)
    d = len(X[0])
    W = zeros(d)

    idx = range(n)
    if rand:
        idx = random.sample(idx, n)
    t = 0
    k = 0
    flag = True
    while True:
        if k == n:
            if flag: break
            k = 0
            flag = True

        i = idx[k]
        if sign(inner(X[i], W)) != Y[i]:
            flag = False
            t += 1
            W = W + alpha * Y[i] * X[i]
        k += 1

    return t


def naive_cycle():
    X, Y = load_data(DATA_FILE)
    t = train(X, Y)
    print t


def predefined_random(n, alpha=1):
    X, Y = load_data(DATA_FILE)
    count = 0
    for i in xrange(n):
        print i
        t = train(X, Y, rand=True, alpha=alpha)
        count += t
    print count/n


def main():
    # question 15
    # naive_cycle()  # 45
    # question 16
    # predefined_random(2000)  # 40
    # question 17
    predefined_random(2000, alpha=0.5)  # 40



if __name__ == '__main__':
    main()

hw1_pocket_perceptron.py

#!/usr/bin/env python
#coding=utf8

"""
Train DATA: https://d396qusza40orc.cloudfront.net/ntumlone%2Fhw1%2Fhw1_18_train.dat
Test DATA:  https://d396qusza40orc.cloudfront.net/ntumlone%2Fhw1%2Fhw1_18_test.dat

Question 18:
    As the test set for "verifying" the g returned by your algorithm (see lecture 4 about verifying).
    The sets are of the same format as the previous one.
    Run the pocket algorithm with a total of 50 updates on D, and verify the performance of w using the test set.
    Please repeat your experiment for 2000 times, each with a different random seed.
    What is the average error rate on the test set?

Question 19:
    Modify your algorithm in Question 18 to return w50 (the PLA vector after 50 updates) instead of
    Wg (the pocket vector) after 50 updates. Run the modified algorithm on D, and verify the performance
    using the test set. Please repeat your experiment for 2000 times, each with a different random seed.
    What is the average error rate on the test set?

Question 20:
    Modify your algorithm in Question 18 to run for 100 updates instead of 50, and verify the performance
    of wPOCKET using the test set. Please repeat your experiment for 2000 times, each with a different random seed.
    What is the average error rate on the test set?
"""

import random
from numpy import array, inner, zeros

TRAIN_FILE = 'pocket_hw1_18_train.dat'
TEST_FILE = 'pocket_hw1_18_test.dat'


def sign(x):
    if x <= 0:
        return -1
    return 1


def load_data(infile):
    X = []
    Y = []
    with open(infile) as f:
        for line in f:
            recs = line.split()
            x = [1] + [float(v) for v in recs[:-1]]
            X.append(tuple(x))
            Y.append(int(recs[-1]))
    return array(X), array(Y)


def test(X, Y, W):
    n = len(Y)
    ne = sum([1 for i in xrange(n) if sign(inner(X[i], W)) != Y[i]])
    error = ne / float(n)
    return error


def train(X, Y, updates=50, pocket=True):
    n = len(Y)
    d = len(X[0])
    W = zeros(d)
    Wg = W

    error = test(X, Y, Wg)
    for j in xrange(updates):
        idx = random.sample(range(n), n)
        for i in idx:
            if sign(inner(X[i], W)) != Y[i]:
                W = W + Y[i] * X[i]
                e = test(X, Y, W)
                if e < error:
                    error = e
                    Wg = W
                break

    if pocket:
        return Wg
    else:
        return W


def main():
    X, Y = load_data(TRAIN_FILE)
    TX, TY = load_data(TEST_FILE)
    error = 0
    n = 200

    for i in xrange(n):
        # question 18, output: 0.1377
        #W = train(X, Y, updates=50)
        # question 19, output: 0.3559
        #W = train(X, Y, updates=50, pocket=False)
        # question 20, output: 0.1141
        W = train(X, Y, updates=100, pocket=True)
        error += test(TX, TY, W)
    print error / n


if __name__ == '__main__':
    main()

NTU - Machine Learning

Codes for Machine Learning Foundations(NTU)
台湾国立大学《机器学习基石》（Coursera版）相关的代码、编程作业等。

课程地址：https://class.coursera.org/ntumlone-001/