def exercise6():
print("\nExercise 6: Testing\n")
perceptron = Perceptron(4, 2, 2)
trainImages = [
np.array([[1, 0], [0, 0]]),
np.array([[0, 0], [0, 1]]),
np.array([[1, 1], [0, 0]]),
np.array([[0, 1], [0, 1]])
]
labels = [1, -1, 1, -1]
perceptron.train(trainImages, labels, 10)
testImages = [
np.array([[1, 0], [0, 0]]),
np.array([[0, 0], [0, 1]]),
np.array([[1, 1], [0, 0]]),
np.array([[0, 1], [0, 1]])
]
expectedResults = [1, -1, 1, -1]
print(testTesting(perceptron, testImages, expectedResults))
testImages = [
np.array([[0, 0], [1, 0]]),
np.array([[0, 0], [1, 1]]),
np.array([[1, 1], [1, 0]]),
np.array([[0, 1], [1, 1]])
]
expectedResults = [0, -1, 1, -1]
print(testTesting(perceptron, testImages, expectedResults))
def fullTest():
trainImages, labels, testImages = loadImages()
per = Perceptron(64, 8, 8)
per.train(trainImages, labels, 10000)
results = per.test(testImages)
print(results)
def exercise5():
print("\nExercise 5: Training\n")
perceptron = Perceptron(4, 2, 2)
images = [np.array([[1, 0], [0, 0]])]
labels = [1]
print(testTraining(perceptron, images, labels))
perceptron = Perceptron(4, 2, 2)
images = [np.array([[1, 0], [0, 0]]), np.array([[0, 0], [0, 1]])]
labels = [1, -1]
print(testTraining(perceptron, images, labels))
perceptron = Perceptron(4, 2, 2)
images = [
np.array([[1, 0], [0, 0]]),
np.array([[0, 0], [0, 1]]),
np.array([[1, 1], [0, 0]])
]
labels = [1, -1, 1]
print(testTraining(perceptron, images, labels))
perceptron = Perceptron(4, 2, 2)
images = [
np.array([[1, 0], [0, 0]]),
np.array([[0, 0], [0, 1]]),
np.array([[1, 1], [0, 0]]),
np.array([[0, 1], [0, 1]])
]
labels = [1, -1, 1, -1]
print(testTraining(perceptron, images, labels))
def testCalcError(expectedResult, labels, results):
perceptron = Perceptron(1, 1, 1)
result = perceptron.calcError(labels, results)
outString = ("Testing with labels={}, results={}: ".format(
labels, results))
if expectedResult != result:
outString += ("{}KO| Expected {} got {}{}").format(
Fore.RED, expectedResult, result, Fore.RESET)
return outString
outString += "{}OK{}".format(Fore.GREEN, Fore.RESET)
return outString
def fullTest():
trainImages, labels, testImages = loadImages()
per = Perceptron(64, 8, 8)
per.train(trainImages, labels, 10000)
results = per.test(testImages)
for image, result in zip(testImages, results):
cv2.imshow("test", image * 255)
print(result)
cv2.waitKey(0)
print(results)
def testForwardPass(network, image, expected):
imageHeight, imageWidth = image.shape[:2]
perceptron = Perceptron(len(network), imageWidth, imageHeight)
perceptron.network = network
result = perceptron.forwardPass(image)
outString = ("Testing with image:\n{}\nnetwork={}: ".format(image, network))
if result != expected:
outString += ("{}KO| Expected {} got {}{}").format(Fore.RED, expected,
result, Fore.RESET)
return outString
outString += "{}OK{}".format(Fore.GREEN, Fore.RESET)
return outString
def testPerceptronInit(numberOfNeurons, imageWidth, imageHeight):
perceptron = Perceptron(numberOfNeurons, imageWidth, imageHeight)
outString = ("Testing with numberOfNeurons={}, imageWidth={}, "
"imageHeight={}: ".format(numberOfNeurons, imageWidth, imageHeight))
if len(perceptron.network) != numberOfNeurons:
outString += ("{}KO| Expected length of network to be {} but "
"got {}{}").format(Fore.RED, len(perceptron.network),
numberOfNeurons, Fore.RESET)
return outString
if type(perceptron.network) != list:
outString += ("{}KO| Expected perceptron.network to be of type {} but "
"got {}{}".format(Fore.RED, list,
type(perceptron.network), Fore.RESET))
return outString
for netIndex, neuron in enumerate(perceptron.network):
if type(neuron) != Neuron:
outString += ("{}KO| Expected all of the objects in "
"perceptron.network to be of type {} but the "
"object at position {} is of type {}{}").format(Fore.RED,
Neuron,
netIndex,
type(neuron),
Fore.RESET)
return outString
for netIndex, neuron in enumerate(perceptron.network):
if neuron.xPos not in range(imageWidth) or neuron.yPos not in range(imageHeight):
outString += ("{}KO| Expected all neuron positions to be within the "
"range [(0, 0), ({}, {})[ but the neuron at index {} "
"has position ({}, {}){}".format(Fore.RED, imageWidth,
imageHeight, netIndex,
neuron.xPos, neuron.yPos,
Fore.RESET))
return outString
neuronPositions = []
for netIndex, neuron in enumerate(perceptron.network):
if (neuron.xPos, neuron.yPos) in neuronPositions:
outString += ("{}KO| Expected all neuron positions to be different "
"but neurons at index {} and neuron at index {} "
"both have position ({}, {}){}".format(Fore.RED,
netIndex,
neuronPositions.index((neuron.xPos, neuron.yPos)),
neuron.xPos,
neuron.yPos,
Fore.RESET))
return outString
else:
neuronPositions.append((neuron.xPos, neuron.yPos))
outString += "{}OK{}".format(Fore.GREEN, Fore.RESET)
return outString
def bonus():
perceptron1 = Perceptron(1,1,1)
perceptron1.network = [Neuron((0,0), 0, False)]
print(testSaveAndLoad(perceptron1))
perceptron1.network = [Neuron((0,0), 0, True)]
print(testSaveAndLoad(perceptron1))
perceptron1.network = [Neuron((0,0), 0, True)]
print(testSaveAndLoad(perceptron1))
perceptron1.network = [Neuron((0,0), 1, True)]
print(testSaveAndLoad(perceptron1))
perceptron2 = Perceptron(2,2,2)
perceptron2.network = [Neuron((0,0), -1, False), Neuron((1,1), 1, False)]
print(testSaveAndLoad(perceptron2))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, False)]
print(testSaveAndLoad(perceptron2))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, True)]
print(testSaveAndLoad(perceptron2))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, True)]
print(testSaveAndLoad(perceptron2))
perceptron2.network = [Neuron((0,0), -1, False), Neuron((1,1), 2, True)]
print(testSaveAndLoad(perceptron2))
perceptron2.network = [Neuron((0,0), 0, True), Neuron((1,1), 1, True)]
print(testSaveAndLoad(perceptron2))
def exercise3():
print("\nExercise 3: Back propagation\n")
perceptron1 = Perceptron(1,1,1)
perceptron1.network = [Neuron((0,0), 0, False)]
expectedNetwork = [Neuron((0,0), 0, False)]
print(testBackProp(perceptron1, expectedNetwork, 1, 0))
perceptron1.network = [Neuron((0,0), 0, True)]
expectedNetwork = [Neuron((0,0), 1, False)]
print(testBackProp(perceptron1, expectedNetwork, 1, 0))
perceptron1.network = [Neuron((0,0), 0, True)]
expectedNetwork = [Neuron((0,0), -1, False)]
print(testBackProp(perceptron1, expectedNetwork, -1, 0))
perceptron1.network = [Neuron((0,0), 1, True)]
expectedNetwork = [Neuron((0,0), 0, False)]
print(testBackProp(perceptron1, expectedNetwork, -1, 1))
perceptron2 = Perceptron(2,2,2)
perceptron2.network = [Neuron((0,0), -1, False), Neuron((1,1), 1, False)]
expectedNetwork = [Neuron((0,0), -1, False), Neuron((1,1), 1, False)]
print(testBackProp(perceptron2, expectedNetwork, -1, 0))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, False)]
expectedNetwork = [Neuron((0,0), -1, False), Neuron((1,1), 1, False)]
print(testBackProp(perceptron2, expectedNetwork, -1, -1))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, True)]
expectedNetwork = [Neuron((0,0), -2, False), Neuron((1,1), 0, False)]
print(testBackProp(perceptron2, expectedNetwork, -1, 0))
perceptron2.network = [Neuron((0,0), -1, True), Neuron((1,1), 1, True)]
expectedNetwork = [Neuron((0,0), 0, False), Neuron((1,1), 2, False)]
print(testBackProp(perceptron2, expectedNetwork, 1, 0))
perceptron2.network = [Neuron((0,0), -1, False), Neuron((1,1), 2, True)]
expectedNetwork = [Neuron((0,0), -1, False), Neuron((1,1), 2, False)]
print(testBackProp(perceptron2, expectedNetwork, 1, 1))
perceptron2.network = [Neuron((0,0), 0, True), Neuron((1,1), 1, True)]
expectedNetwork = [Neuron((0,0), -1, False), Neuron((1,1), 0, False)]
print(testBackProp(perceptron2, expectedNetwork, -1, 1))