def cross_validation(method, k=5):
if method == "lda":
Classifier = LDAClassifier
elif method == "qda":
Classifier = QDAClassifier
else:
raise Exception("lda or qda only")
timer.start("folding data into", k, "copies")
data_slice = [None] * k
labl_slice = [None] * k
train_rate = [0.0] * k
valid_rate = [0.0] * k
n = len(labl)
m = n / k
for i in range(k):
data_slice[i] = data[(i * m):min((i + 1) * m, n)]
labl_slice[i] = labl[(i * m):min((i + 1) * m, n)]
timer.end("done")
for j in range(k):
timer.start("validation iteration #", j)
training_data = np.concatenate(
tuple(data_slice[i] for i in range(k) if i != j))
training_labl = np.concatenate(
tuple(labl_slice[i] for i in range(k) if i != j))
print ".... data formating done"
c = LDAClassifier(training_data, training_labl)
print ".... classifier training done"
train_rate[j] = c.score(c.classify_all(training_data), training_labl)
print ".... training accuracy computation done"
valid_rate[j] = c.score(c.classify_all(data_slice[j]), labl_slice[j])
print ".... validation accuracy computation done"
timer.end("done; training accuracy =", train_rate[j],
"; validation accuracy =", valid_rate[j])
print k, "fold cross validation for", method, "on dataset", which, "complete"
print ".... overall training accuracy =", np.mean(train_rate)
print ".... overall validation accuracy =", np.mean(valid_rate)
"""
Problem :
How many distinct terms are in the sequence generated by a^b for 2 ≤ a ≤ 100
and 2 ≤ b ≤ 100?
Performance time: ~0.0059s
"""
from timer import timer
timer.start()
print(len(set(a**b for a in range(2, 101) for b in range(2, 101))))
timer.stop()
"""
Problem:
Find the sum of digits in the numerator of the 100th convergent of the
continued fraction for e.
Performance time: ~0.0012s
"""
from fractions import Fraction
from timer import timer
from utils import sum_of_digits
timer.start()
a = [int(n / 3 * 2) if n % 3 == 0 else 1 for n in range(1, 101)]
a[0] = 2
def convergent_e(limit=-1, index=0):
if limit == 0:
return a[0]
elif index == limit - 1:
return a[index] + Fraction(1, a[index+1])
else:
return a[index] + Fraction(1, convergent_e(limit, index+1))
print(sum_of_digits(Fraction(convergent_e(99)).numerator))
timer.stop()
import numpy as np
from os import path
from scipy.io import loadmat
from timer import timer
from classifier import LDAClassifier, QDAClassifier
""" TODO: choose either mnist or spam >>HERE<< """
which = "spam.mat"
which = "mnist.mat"
""" TODO: choose either mnist or spam >>HERE<< """
timer.start("reading", which, "data from matlab file")
raw = loadmat(path.join(path.dirname(__file__), "data", which))
raw_data = raw['data']
raw_labl = raw['label'][0]
timer.end("done")
timer.start("permuting data randomly")
np.random.seed(0)
ordering = np.random.permutation(len(raw_data))
data = np.ndarray(shape=raw_data.shape, dtype=raw_data.dtype)
labl = np.ndarray(shape=raw_labl.shape, dtype=raw_labl.dtype)
for old, new in enumerate(ordering):
data[new] = raw_data[old]
labl[new] = raw_labl[old]
del raw, raw_data, raw_labl, ordering
timer.end("done")
def cross_validation(method, k=5):
if method == "lda":
Classifier = LDAClassifier