Data Source: http://archive.ics.uci.edu/ml/datasets/Optical+Recognition+of+Handwritten+Digits
Number of Instances: 5620
Number of Attributes: 64
Relevant Information (from website):
We used preprocessing programs made available by NIST to extract normalized bitmaps of handwritten digits from a preprinted form. From a total of 43 people, 30 contributed to the training set and different 13 to the test set. 32x32 bitmaps are divided into nonoverlapping blocks of 4x4 and the number of on pixels are counted in each block. This generates an input matrix of 8x8 where each element is an integer in the range 0..16. This reduces dimensionality and gives invariance to small distortions.
Number of Instances: 5620
Number of Attributes: 64
Relevant Information (from website):
We used preprocessing programs made available by NIST to extract normalized bitmaps of handwritten digits from a preprinted form. From a total of 43 people, 30 contributed to the training set and different 13 to the test set. 32x32 bitmaps are divided into nonoverlapping blocks of 4x4 and the number of on pixels are counted in each block. This generates an input matrix of 8x8 where each element is an integer in the range 0..16. This reduces dimensionality and gives invariance to small distortions.
In [61]:
import os, sys, sklearn
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm
%pylab inline
1. Load and look at the data¶
In [9]:
!ls OpticalDigits
In [104]:
filepath_train = 'OpticalDigits/optdigits.tra'
filepath_test = 'OpticalDigits/optdigits.tes'
In [106]:
data_train = pd.DataFrame.from_csv(filepath_train, header = None, index_col = None)
data_test = pd.DataFrame.from_csv(filepath_test, header = None, index_col = None)
In [107]:
print "Size of training set: %d"%len(data_train)
print "Size of test set: %d"%len(data_test)
In [108]:
#split the data into features and labels
train_x = data_train[data_train.columns[:64]]
train_y = data_train[data_train.columns[64]]
test_x = data_test[data_test.columns[:64]]
test_y = data_test[data_test.columns[64]]
In [109]:
#verify that the data is correct
for i in range(5):
print "Label: %d"%train_y[i]
img = plt.figure()
plt.imshow(train_x.loc[i].reshape((8,8)), cmap = cm.Greys_r)
img.set_size_inches ((0.8,0.8))
plt.show()
In [110]:
#look at the label distribution of the data
train_y.hist()
Out[110]:
2. Train a simple model on the data¶
Nearest Neighbor (kNN, k=1)¶
In [111]:
# Reference: http://cs231n.github.io/classification/
import numpy as np
class NearestNeighbor:
def __init__(self):
pass
def train(self, X, y):
""" X is N x D where each row is an example. Y is 1-dimension of size N """
# the nearest neighbor classifier simply remembers all the training data
self.Xtr = X
self.ytr = y
def predict(self, X):
""" X is N x D where each row is an example we wish to predict label for """
num_test = X.shape[0]
# lets make sure that the output type matches the input type
Ypred = np.zeros(num_test, dtype = self.ytr.dtype)
# loop over all test rows
for i in xrange(num_test):
# find the nearest training image to the i'th test image
# using the L1 distance (sum of absolute value differences)
distances = np.sum(np.abs(self.Xtr - X[i,:]), axis = 1)
min_index = np.argmin(distances) # get the index with smallest distance
Ypred[i] = self.ytr[min_index] # predict the label of the nearest example
return Ypred
In [119]:
knn = NearestNeighbor()
knn.train(train_x.as_matrix(), train_y.as_matrix())
In [140]:
test_x.head(5)
Out[140]:
In [131]:
print "Predicting %d examples..."%len(test_x)
%time pred = knn.predict(test_x.as_matrix())
In [132]:
import sklearn.metrics as metrics
print "Accuracy: %.2f"%metrics.accuracy_score(test_y, pred)
In [133]:
print metrics.classification_report(test_y, pred)
In [137]:
print metrics.confusion_matrix(test_y, pred)
print "X-axis: predicted, Y-axis: true"
In [157]:
from sklearn import svm
clf = svm.SVC(decision_function_shape='ovo')
clf.fit(train_x.as_matrix(), train_y)
Out[157]:
In [158]:
dec = clf.predict(test_x)
In [160]:
print metrics.confusion_matrix(test_y, dec)
print "Accuracy: %.2f"%metrics.accuracy_score(test_y, dec)
In []: