def cee_accuracy():
from hw6code import ERROR_CEE, Neural_Network
import random
np.random.seed(12)
random.seed(13)
data = load_trains()
trains = data[:-10000]
tests = data[-5000:]
nnetc = Neural_Network(error_function=ERROR_CEE)
nnetc.progressive_init(trains)
def get_accuracy():
total = 0
correct = 0
for test in tests:
hw = nnetc.predict(test)
hj = vector_to_label(hw)
if hj == test.yl:
correct += 1
total += 1
return correct / total
for reps_goal in xrange(0, 300000, 1000):
nnetc.reps = reps_goal
nnetc.prog_train()
accuracyc = get_accuracy()
print reps_goal, accuracyc
ceelist.append(accuracyc)
def training_accuracy():
from hw6code import Neural_Network, ERROR_CEE, ERROR_MSE
import random
np.random.seed(12)
random.seed(13)
data = load_trains()
nnet = Neural_Network(error_function=ERROR_CEE)
nnet.train(data)
total = 0
correct = 0
for test in data:
hw = nnet.predict(test)
hj = vector_to_label(hw)
if hj == test.yl:
correct += 1
total += 1
print correct / total
def kaggle():
from hw6code import Neural_Network
import random
import csv
np.random.seed(12)
random.seed(13)
trains = load_trains()
tests = load_tests()
nnet = Neural_Network()
nnet.train(trains)
count = 1
print("WRITING")
with open('hw6submit.csv', 'w') as f:
digitwriter = csv.writer(f, delimiter=',')
digitwriter.writerow(['Id', 'Category'])
for test in tests:
hv = nnet.predict(test)
hl = vector_to_label(hv)
digitwriter.writerow([count, hl])
count += 1