Y_test = None
print("Non 0-1 labels removed from testing dataset!")
print("\nTraining SVM on MNIST dataset...")
svm = SupportVectorMachine()
svm.train(X, Y, 1)
print("SVM trained!")
print("\nTraining Linear Regression on MNIST dataset...")
linear = LinearRegression()
linear.train(X, Y)
print("Linear regression trained!")
print("\nTraining Logistic Regression on MNIST dataset...")
logistic = LogisticRegression()
logistic.train(X, Y)
print("Logistic regression trained!")
# Test SVM
print("\nRunning SVM on test data...")
misclassified = svm.test(X2, Y2)
print("Generalization Error:", round(misclassified/Y2.size, 3))
print("Misclassified:", misclassified, "/", Y2.size)
print("Accuracy (on test data):", round((1 - (misclassified/Y2.size)) * 100, 3), '%')
# Test Linear Regression
print("\nRunning Linear Regression on test data...")
misclassified = linear.test(X2, Y2)
print("Generalization Error:", round(misclassified/Y2.size, 3))
print("Misclassified:", misclassified, "/", Y2.size)
print("Accuracy (on test data):", round((1 - (misclassified/Y2.size)) * 100, 3), '%')
from mnist import MNIST
mndata = MNIST('./MNIST')
trImg, trLab = mndata.load_training()
teImg, teLab = mndata.load_testing()
trImg = np.asanyarray(trImg)
trLab = np.asanyarray(trLab)
teImg = np.asanyarray(teImg)
teLab = np.asanyarray(teLab)
usps = LoadUSPS.LoadUSPS('proj3_images.zip')
uspsImg, uspsLab = usps.load()
#1> logistic Regression
logistic = LogisticRegression(28 * 28, 10)
logistic.train(trImg, trLab, lr = 0.3)
accuracy = logistic.test(teImg, teLab)
uspsacc = logistic.test(uspsImg, uspsLab)
print('logisticregression accuracy :', accuracy, uspsacc)
#grid search for best learning rate performance
#for lr in [0.5, 0.3, 0.1, 0.05, 0.01]:
# logistic.train(trImg, trLab, lr = 0.1)
# accuracy = logistic.test(teImg, teLab)
# print(lr, accuracy)
#2> Multilayer perceptron implementation using tensorflow
mlp = MLP.MLP()
mlp.train()
# -*- coding: utf-8 -*- """ Created on Mon Sep 4 17:29:50 2017 @author: heman """ from request_data_link import get import numpy as np from logistic import LogisticRegression link = 'http://data.princeton.edu/wws509/datasets/copen.raw' m, n, parsed_data = get(link, 6) index = list(range(0, parsed_data.size, 6)) parsed_data = np.delete(parsed_data, index) index = list(range(4, parsed_data.size, 5)) targets = parsed_data[index] data = np.delete(parsed_data, index).reshape(m, n - 1) del link, m, n, parsed_data, index regr = LogisticRegression() regr.train(data, targets, iter=1000000, step=0.001, lamda=0) labels, predictions = regr.test(data, targets)
