def main():
"""This is the driver program for showing results from part 1.
Uncomment the code below if you need to pickle a new trained
perceptron."""
digitsMatrices, digitsTargets = loadPatterns("digits_train.txt")
digitsTests, digitsTestTargets = loadPatterns("digits_test.txt")
target = np.array(())
for p in range(2500):
arrayOfTen = np.array(())
beginning = np.append(np.zeros(digitsTargets[p]), np.ones(1))
arrayOfTen = np.append(beginning, np.zeros(9 - digitsTargets[p]))
target = np.append(target, arrayOfTen)
target = np.reshape(target, (2500, 10))
"""trainEta = 1 #Change only these when tweaking for better results
numberOfPatterns = 10000 #Change only these when tweaking for better results
print("Initialized Perceptron: \n")
digitsPerceptron = Perceptron(196,10,3,.1)
digitsPerceptron.train(digitsMatrices,target,numberOfPatterns,trainEta)
digitsPerceptron.save("part3.dat")"""
newPerceptron = Perceptron(196, 10, 3, .1)
unPickled = newPerceptron.load("part3.dat")
confusion = np.zeros((10, 10), "int")
for k in range(2500):
row = k // 250
col = np.argmax(unPickled.test(digitsTests[k]))
confusion[row, col] += 1
print(confusion)
"""threshold = .23 #Change only these when tweaking for better results
def main():
"""This is the driver program for showing results from part 1.
Uncomment the code below if you need to pickle a new trained
perceptron."""
digitsMatrices, digitsTargets = loadPatterns("digits_train.txt")
digitsTests, digitsTestTargets = loadPatterns("digits_test.txt")
target = np.array(())
for p in range(2500):
arrayOfTen = np.array(())
beginning = np.append(np.zeros(digitsTargets[p]),np.ones(1))
arrayOfTen = np.append(beginning,np.zeros(9-digitsTargets[p]))
target = np.append(target,arrayOfTen)
target = np.reshape(target,(2500,10))
"""trainEta = 1 #Change only these when tweaking for better results
numberOfPatterns = 10000 #Change only these when tweaking for better results
print("Initialized Perceptron: \n")
digitsPerceptron = Perceptron(196,10,3,.1)
digitsPerceptron.train(digitsMatrices,target,numberOfPatterns,trainEta)
digitsPerceptron.save("part3.dat")"""
newPerceptron = Perceptron(196,10,3,.1)
unPickled = newPerceptron.load("part3.dat")
confusion = np.zeros((10,10),"int")
for k in range(2500):
row = k // 250
col = np.argmax(unPickled.test(digitsTests[k]))
confusion[row,col] += 1
print(confusion)
"""threshold = .23 #Change only these when tweaking for better results
def main():
"""This is the driver program for showing results from part 1.
Uncomment the code below if you need to pickle a new trained
perceptron."""
twoMatrices, twoTargets = loadPatterns("digits_train.txt")
twoTests, twoTestsTargets = loadPatterns("digits_test.txt")
target = np.zeros(2500)
target[500:750] = 1
falsePositives = 0
misses = 0
"""
trainEta = 1 #Change only these when tweaking for better results
numberOfPatterns = 2000 #Chang only these when tweaking for better results
print("Initialized Perceptron: \n")
twoPerceptron = Perceptron(196,1,3,.1)
twoPerceptron.train(twoMatrices,target,numberOfPatterns,trainEta)
twoPerceptron.save()"""
newPerceptron = Perceptron(196,1,3,.1)
unPickled = newPerceptron.load('weights.dat')
threshold = .23 #Change only these when tweaking for better results
#print("Expect greater than " + str(threshold) +": " + str(unPickled.test(twoTests[521])))
for i in range(len(twoTests)):
test = unPickled.test(twoTests[i])
if test >= threshold and target[i] == 0:
falsePositives += 1
if test < threshold and target[i] == 1:
misses += 1
print("Misses: " + str(misses) + "/250")
print("False Positives: " + str(falsePositives) + "/2250")
def main():
"""This is the driver program for showing results from part 1.
Uncomment the code below if you need to pickle a new trained
perceptron."""
twoMatrices, twoTargets = loadPatterns("digits_train.txt")
twoTests, twoTestsTargets = loadPatterns("digits_test.txt")
target = np.zeros(2500)
target[500:750] = 1
falsePositives = 0
misses = 0
"""
trainEta = 1 #Change only these when tweaking for better results
numberOfPatterns = 2000 #Chang only these when tweaking for better results
print("Initialized Perceptron: \n")
twoPerceptron = Perceptron(196,1,3,.1)
twoPerceptron.train(twoMatrices,target,numberOfPatterns,trainEta)
twoPerceptron.save()"""
newPerceptron = Perceptron(196, 1, 3, .1)
unPickled = newPerceptron.load('weights.dat')
threshold = .23 #Change only these when tweaking for better results
#print("Expect greater than " + str(threshold) +": " + str(unPickled.test(twoTests[521])))
for i in range(len(twoTests)):
test = unPickled.test(twoTests[i])
if test >= threshold and target[i] == 0:
falsePositives += 1
if test < threshold and target[i] == 1:
misses += 1
print("Misses: " + str(misses) + "/250")
print("False Positives: " + str(falsePositives) + "/2250")
def main():
"""This is the driver program for showing results from part 1."""
inputArray = np.array([[0,0],[0,1],[1,0],[1,1]])
targetXOR = np.array([[0],[1],[1],[0]])
print("Initializing 1X3 Perceptron....")
testP = Perceptron(2,1,3)
print(str(testP) + "\n")
print("\n"+"Testing learning XOR: \n")
testP.train(inputArray, targetXOR)
print("Training results:\n")
print("Expect Close to 0: " + str(testP.test(np.array([0,0]))))
print("Expect Close to 1: " + str(testP.test(np.array([0,1]))))
print("Expect Close to 1: " + str(testP.test(np.array([1,0]))))
print("Expect Close to 0: " + str(testP.test(np.array([1,1]))) + "\n")
def main():
"""This is the driver program for showing results from part 1."""
inputArray = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
targetXOR = np.array([[0], [1], [1], [0]])
print("Initializing 1X3 Perceptron....")
testP = Perceptron(2, 1, 3)
print(str(testP) + "\n")
print("\n" + "Testing learning XOR: \n")
testP.train(inputArray, targetXOR)
print("Training results:\n")
print("Expect Close to 0: " + str(testP.test(np.array([0, 0]))))
print("Expect Close to 1: " + str(testP.test(np.array([0, 1]))))
print("Expect Close to 1: " + str(testP.test(np.array([1, 0]))))
print("Expect Close to 0: " + str(testP.test(np.array([1, 1]))) + "\n")
"""
from backprop import Perceptron
import numpy as np
if __name__ == '__main__':
print("Assignment 3 : Part 1")
boolInputs = np.vstack([np.array([0,0]), np.array([0,1]),
np.array([1,0]), np.array([1,1])])
xorP = Perceptron(2,1,3,1)
xorOutputs = np.vstack([np.array([0]), np.array([1]), np.array([1]),
np.array([0])])
xorP.train(boolInputs, xorOutputs, 10000)
print("\nTESTING NEURAL NET ON XOR")
print("NITER = 10000; ETA = 1; HIDDEN UNITS = 3")
xorTest = xorP.test(boolInputs)
for i in range(len(xorTest)):
print("Input = " + str(boolInputs[i]) + " Output = " + str(xorTest[i]))
#andP = Perceptron(2,1,3)
#andOutputs = np.vstack([np.array([0]), np.array([0]), np.array([0]),
from digits import *
if __name__ == '__main__':
#p = Perceptron(196, 10, 20, .5)
inputArray, visualArray, outputArray = readPattern("digits_train.txt")
inputArray2, visualArray2, outputArray2 = readPattern("digits_test.txt")
confusionMatrix = np.zeros((10,10))
print("TESTING PERCEPTRON: CLASSIFYING ALL 10 DIGITS")
print("NITER = 5000; ETA = .75; HIDDEN UNITS = 20")
#p.train(inputArray, outputArray, 5000, .75)
#p.save("part3perceptron.txt")
p = Perceptron(1,1,1,1)
p.load("part3perceptron.txt")
testOutput = p.test(inputArray2)
for j in range(len(outputArray)):
a = np.argmax(outputArray[j])
b = np.argmax(testOutput[j])
confusionMatrix[a][b] += 1
print(confusionMatrix)
"""
Stephen Rowley
roc.py
Plots a Receiver Operator Characteristic for a backprop perceptron network
Professor Levy
"""
from backprop import Perceptron
import numpy as np
from digits import *
import matplotlib.pyplot as plt
#load in perceptron
p = Perceptron(1,1,1,1)
p.load("part2perceptron.txt")
#initialize
falsePosList = np.array([])
missList = np.array([])
truePosList = np.array([])
inputArray2, visualArray2, outputArray2 = readPattern("digits_test.txt")
xAxis = np.arange(0,1,.05)
if (outputArray2[1][2] == 0):
newOutputArray = np.array([0])
if (outputArray2[1][2] == 1):
newOutputArray = np.array([1])
for i in range(len(outputArray2)-1):
if (outputArray2[i+1][2] == 1):
newOutputArray = np.vstack([newOutputArray, 1])