Learning mod 2 (xor) using binary representation

felipe_test.py

from sklearn import grid_search
from sklearn import metrics
from sklearn import tree

import numpy as np

def create_sample_decimal(num_rows, max_val):
    X = np.random.randint(0, max_val, size=(num_rows, 2))
    y = (X[:, 0] + X[:, 1]) % 2

    return X, y

def create_sample_binary(num_rows, max_val):
    first_col = np.random.randint(0, max_val, size=num_rows)
    second_col = np.random.randint(0, max_val, size=num_rows)

    first_col = np.asanyarray(first_col, dtype=np.uint8)
    second_col = np.asanyarray(second_col, dtype=np.uint8)

    bin_first = np.unpackbits(first_col)
    bin_second = np.unpackbits(second_col)

    bin_first = np.reshape(bin_first, (num_rows, 8))
    bin_second = np.reshape(bin_second, (num_rows, 8))

    #simple sanity check for the reshape above
    assert (bin_first[0] == np.unpackbits(first_col[0])).all()

    X = np.hstack((bin_first, bin_second))
    y = (first_col + second_col) % 2
    return X, y

def main():
    n_train = 1000
    n_test = 100
    max_val = 255
    use_bin = True

    if use_bin:
        X_train, y_train = create_sample_binary(n_train, max_val)
        X_test, y_test = create_sample_binary(n_test, max_val)
    else:
        X_train, y_train = create_sample_decimal(n_train, max_val)
        X_test, y_test = create_sample_decimal(n_test, max_val)

    clf = tree.DecisionTreeClassifier('entropy')
    estimator = clf.fit(X_train, y_train)
    y_predict = estimator.predict(X_test)
    
    print metrics.classification_report(y_test, y_predict)

if __name__ == '__main__':
    main()