def obtainKaggleTestResults(Theta1, Theta2):
''' Apply the learned neural network paramaters to predict the
results on the Kaggle Test Data.
'''
# get the data to do the final test.
m, n, X = getKaggleTestData()
# Select some random images from X
print('Selecting random examples of the test data to display.\n')
sel = np.random.permutation(m)
sel = sel[0:100]
images = X[sel, :]
# display the sample images
displayData(images)
# predict the labels
pred = predict(Theta1, Theta2, X)
# print out what the images are predicted to be
print('Here are the predicted labels for these random images: \n')
print(pred[sel].reshape(10, 10))
# Pause
print("Program paused. Press Ctrl-D to continue.\n")
code.interact(local=dict(globals(), **locals()))
print(" ... continuing\n ")
return pred
def displaySampleData(X, y):
''' Display random images as a grid, along with the corresponding labels'''
m, n = X.shape
# Select some random images from X
print('Selecting random examples of the data to display.\n')
sel = np.random.permutation(m)
sel = sel[0:100]
images = X[sel, :]
# display the sample images
displayData(images)
# Print Out the labels for what is being seen.
print('These are the labels for the data ...\n')
print(y[sel, :].reshape(10, 10))
def visualizeNN(Theta, hidden_layer_size, iters, _lambda):
""" Visualize NN Weights that have been learned.
You can now "visualize" what the neural network is learning by
displaying the hidden units to see what features they are capturing in
the data.
"""
print ("\nVisualizing Trained Neural Network... \n")
print (
"\n Parms: Hidden Layer Units: {0} Max Iters: {1} Lambda: {2} \n".format(hidden_layer_size, iters, _lambda)
)
displayData(Theta[:, 1:])
# Pause
print ("Program paused. Press Ctrl-D to continue.\n")
code.interact(local=dict(globals(), **locals()))
print (" ... continuing\n ")
return
test = pd.read_csv(fnTest)
M = test.as_matrix() # gives numpy array
X = M[:,0:] # The test data
# Scale the input grey scale pixels by 255 to between -1 to 1 if scale by 1
#
X = X.astype('float')
X = (X - ( X.max() /float(scale)))/X.max()
# Find number of rows examples "m", & columns features "n"
m, n = X.shape
return m, n, X
'''
# old matlab code below
% There is no label y,so comment out this part.
% y = X(:, 1);
% Replace the 0's with 10 in y for indexing purposes
% y(y==0) = 10;
% X = X(:, 2:end);
m = size(TEST, 1);
% Re-orient the images
for i = 1: m
TEST(i, :) = reshape(reshape(TEST(i, :), ...
image_size, image_size)', ...
1, input_layer_size);
end