Created
November 6, 2017 12:03
-
-
Save farizrahman4u/f51fabd8d2e2539ef7343becb36be95c to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
from simple_classifier import SimpleClassifier | |
def vectorize(x): | |
# vectorize a string | |
if len(x) > 1: | |
return np.sum([vectorize(c) for c in x], axis=0) | |
if x == '.': | |
i = 27 | |
elif x == ' ': | |
i = 26 | |
else: | |
x = x.lower() | |
i = ord(x) - 97 | |
oh = np.zeros(28) | |
oh[i] = 1 | |
return oh | |
def decode(x): | |
# decode a one hot to a character | |
i = np.argmax(x) | |
if i == 27: | |
return '.' | |
if i == 26: | |
return ' ' | |
return chr(i + 97) | |
# the data | |
X = ['hi', 'his', 'hise', 'hisen', 'isend', 'send ', 'end n', 'nd nu', 'd nud', ' nude', 'nudes'] | |
Y = ['s', 'e', 'n', 'd', ' ', 'n', 'u', 'd', 'e', 's', '.'] | |
# vectorize data | |
X = np.array(list(map(vectorize, X))) | |
Y = np.array(list(map(vectorize, Y))) | |
clf = SimpleClassifier(28, 28, num_pop=50) | |
clf.fit(X, Y, epochs=1000, batch_size=11, validation_data=(X, Y)) | |
assert clf.evaluate(X, Y)['accuracy'] == 1 | |
# compress weights | |
# remove unnecessary weights to obtain sparse matrix | |
sparse_matrix = [] | |
w = clf.weights[0] | |
for i in range(28): | |
for j in range(28): | |
x = w[i, j] | |
w[i, j] = 0 | |
if clf.evaluate(X, Y)['accuracy'] < 1: | |
w[i, j] = x | |
sparse_matrix.append((i, j, float(str(np.around(x, 1))))) | |
print(sparse_matrix) | |
# test | |
w = np.zeros((28, 28)) | |
for r in sparse_matrix: | |
w[r[0], r[1]] = r[2] | |
clf.weights = [w] | |
print(clf.evaluate(X, Y)) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment