def run_experiment2(filename):
training_data = common.init_data(common.constants.TRAINING_SIZE,
common.constants.DATA_DIM + 1)
test_data = common.init_data(common.constants.TEST_SIZE,
common.constants.DATA_DIM + 1)
gold_data = [0 for x in range(common.constants.TEST_SIZE)]
print("Linear Classifier : Dataset 1")
#generating data should be hidden from students!
read_data(training_data, test_data, gold_data, filename)
#this is one of the two student functions
student_code.part_two_classifier(training_data, test_data)
#part 1 grading
error = 0
for i in range(common.constants.TEST_SIZE):
if (test_data[i][common.constants.DATA_DIM] != gold_data[i]):
error += 1
print("Incorrect classificattions is " + str(error) + " out of " +
str(common.constants.TEST_SIZE))
success = True
if (error <= float(common.constants.TEST_SIZE) * .05):
print("(" + bcolors.GREEN + "SUCCESS" + bcolors.NORMAL + ")")
else:
success = False
print("(" + bcolors.RED + "FAIL" + bcolors.NORMAL + ") maximum " +
str(float(common.constants.TEST_SIZE) * .05))
print
return success
def part_two_classifier(data_train, data_test):
# PUT YOUR CODE HERE
# Access the training data using "data_train[i][j]"
# Training data contains 3 cols per row: X in
# index 0, Y in index 1 and Class in index 2
# Access the test data using "data_test[i][j]"
# Test data contains 2 cols per row: X in
# index 0 and Y in index 1, and a blank space in index 2
# to be filled with class
# The class value could be a 0 or a 8
w = [[0, 0] for x in range(10)]
all_labeled = False
c = common.init_data(common.constants.TRAINING_SIZE, 1)
p = common.init_data(common.constants.TRAINING_SIZE, 1)
while all_labeled != True:
i = 0
# print(i)
for row in data_train:
# print(i)
# print(f"data: {row}, weight: {w}")
c[i] = multi_classify_train(row, w)
# print(f"class = {c}")
if c[i] != row[2]:
w[c[i]][0] -= (row[0]) * (1 / 100)
w[c[i]][1] -= (row[1]) * (1 / 100)
w[int(row[2])][0] += (row[0]) * (1 / 100)
w[int(row[2])][1] += (row[1]) * (1 / 100)
i += 1
# print(f"checking trained labelling")
for i in range(len(data_train)):
p[i] = data_train[i][2]
if check_all_labels(c, p):
print(f"all classes are labeled correctly")
all_labeled = True
i = 0
for row in data_test:
data_test[i][2] = multi_classify_train(row, w)
i += 1
return