Learning Data Mining with Python

Beating CAPTCHAs with Neural Networks
Finally, we will create our dataset. This dataset array is three-dimensional, as it
is an array of two-dimensional images. Our classifier will need a two-dimensional
array, so we simply flatten the last two dimensions:
X = dataset.reshape((dataset.shape[0], dataset.shape[1] * dataset.
shape[2]))
Finally, using the train_test_split function of scikit-learn, we create a set of data
for training and one for testing. The code is as follows:
from sklearn.cross_validation import train_test_split
X_train, X_test, y_train, y_test = \
train_test_split(X, y, train_size=0.9)
Training and classifying
We are now going to build a neural network that will take an image as input and
try to predict which (single) letter is in the image.
We will use the training set of single letters we created earlier. The dataset itself is
quite simple. We have a 20 by 20 pixel image, each pixel 1 (black) or 0 (white). These
represent the 400 features that we will use as inputs into the neural network. The
outputs will be 26 values between 0 and 1, where higher values indicate a higher
likelihood that the associated letter (the first neuron is A, the second is B, and so
on) is the letter represented by the input image.
We are going to use the PyBrain library for our neural network.
As with all the libraries we have seen so far, PyBrain can be installed
from pip: pip install pybrain.
The PyBrain library uses its own dataset format, but luckily it isn't too difficult to
create training and testing datasets using this format. The code is as follows:
from pybrain.datasets import SupervisedDataSet
First, we iterate over our training dataset and add each as a sample into a new
SupervisedDataSet instance. The code is as follows:
training = SupervisedDataSet(X.shape[1], y.shape[1])
for i in range(X_train.shape[0]):
training.addSample(X_train[i], y_train[i])
[ 172 ]