print '#####################'
print 'Current condition: '
print '- hidden architecture', condition[0]
print '- regularization rate', condition[1]
print '- batch size', condition[2]
print '- Epoch number', condition[3]
# 10-fold
result_array = numpy.zeros((10, 3))
for i in xrange(10):
print (('Experiment %d/10 is on-going') % (i+1))
datasets = load10FeatureData(input_data, target_data, idx, i)
# test_result = [training_accuracy, testing_accuracy, f-score]
test_result = test_mlp(architecture = condition[0],
reg_rate = condition[1],
batch_size = condition[2],
epoch_num = condition[3],
datasets = datasets)
result_array[i,:] = numpy.array(test_result)
result_mean = numpy.mean(result_array, axis=0)
result_std = numpy.std(result_array, axis=0)
result = numpy.concatenate((result_mean, result_std), axis=1)
cond_result_list.append(result)
print 'Result: '
print (('Mean training accuracy %f %%, testing accuracy %f %%, f-score %f') % (result_mean[0], result_mean[1], result_mean[2]))
end_time = time.clock()
print (('It takes for %.2fm') % ((end_time - start_time)/60.))
# Output cond_result
# Ubuntu
input_data = numpy.loadtxt(open('/home/heehwan/Documents/workspace/data/DBN_data/X_1405_10features.csv', 'rb'), delimiter=',')
target_data = numpy.loadtxt(open('/home/heehwan/Documents/workspace/data/DBN_data/Y_1405_10features.csv', 'rb'), delimiter=',')
# Make dataset index
idx = range(750)
random.seed(1)
random.shuffle(idx)
print '... starting experiments'
start_time = time.clock()
# 10-fold
result_array = numpy.zeros((10, 4))
for i in xrange(10):
print (('Experiment %d/10 is on-going') % (i+1))
datasets = load10FeatureData(input_data, target_data, idx, i)
# test_result = [training_accuracy, f-score, testing_accuracy, f-score]
test_result = test_mlp(architecture = arch,
reg_rate = reg,
batch_size = batch_size,
epoch_num = epoch,
datasets = datasets)
result_array[i,:] = numpy.array(test_result)
end_time = time.clock()
print (('It takes for %.2fm') % ((end_time - start_time)/60.))
# Result
pickle.dump(result_array, open('result_400_3_v1.pkl', 'wb'))
print result_array
'/home/heehwan/Documents/workspace/data/DBN_data/Y_1405_10features.csv',
'rb'),
delimiter=',')
# Make dataset index
idx = range(750)
random.seed(1)
random.shuffle(idx)
print '... starting experiments'
start_time = time.clock()
# 10-fold
result_array = numpy.zeros((10, 4))
for i in xrange(10):
print(('Experiment %d/10 is on-going') % (i + 1))
datasets = load10FeatureData(input_data, target_data, idx, i)
# test_result = [training_accuracy, f-score, testing_accuracy, f-score]
test_result = test_mlp(architecture=arch,
reg_rate=reg,
batch_size=batch_size,
epoch_num=epoch,
datasets=datasets)
result_array[i, :] = numpy.array(test_result)
end_time = time.clock()
print(('It takes for %.2fm') % ((end_time - start_time) / 60.))
# Result
pickle.dump(result_array, open('result_400_3_v1.pkl', 'wb'))
print result_array