def five_fold(): #7c
errors = []
iteration_count = 10
sub_iteration_count = 5
for i in xrange(iteration_count):
samples_count = 500
testing_samples_count = 100
training_samples_count = samples_count - testing_samples_count
lmda = np.exp(-2)
x, y, v = data_generator2(samples_count, gen_coefficients())
permuted = np.random.permutation(zip(x.T, y))
mse = 0
for j in xrange(sub_iteration_count):
testXtrans, testY = zip(*permuted[j*testing_samples_count:(j+1)*testing_samples_count])
testX = np.array(testXtrans).T
testY = np.array(testY)
permutedTrain = np.append(permuted[0:j*testing_samples_count], permuted[(j+1)*testing_samples_count:], axis=0)
trainXtrans, trainY = zip(*permutedTrain)
trainX = np.array(trainXtrans).T
trainY = np.array(trainY)
mse += calculate_MSE(trainX, trainY, testX, testY, lmda) / sub_iteration_count #7d
errors.append(mse)
print "Five fold average MSE: ", sum(errors)/len(errors) #7e
print "Five fold variance: ", np.var(errors)
def holdout():
errors = []
iteration_count = 10 #7a
for i in xrange(iteration_count):
samples_count = 500
testing_samples_count = 100
training_samples_count = samples_count - testing_samples_count
lmda = np.exp(-2)
x, y, v = data_generator2(samples_count, gen_coefficients())
permuted = np.random.permutation(zip(x.T, y))
testXtrans, testY = zip(*permuted[0:testing_samples_count])
testX = np.array(testXtrans).T
testY = np.array(testY)
trainXtrans, trainY = zip(*permuted[testing_samples_count:])
trainX = np.array(trainXtrans).T
trainY = np.array(trainY)
mse = calculate_MSE(trainX, trainY, testX, testY, lmda)
errors.append(mse)
print "Holdout average MSE: ", sum(errors)/len(errors) #7b
print "Holdout variance: ", np.var(errors)
