如何使Keras神经网络优于Iris数据的Logistic回归

Question

I am comparing Keras Neural-Net with simple Logistic Regression from Scikit-learn on IRIS data. I expect that Keras-NN will perform better, as suggested by this post .

But why by mimicking the code there, the result of Keras-NN is lower than Logistic regression?

import seaborn as sns
import numpy as np
from sklearn.cross_validation import train_test_split
from sklearn.linear_model import LogisticRegressionCV
from keras.models import Sequential
from keras.layers.core import Dense, Activation
from keras.utils import np_utils

# Prepare data
iris = sns.load_dataset("iris")
X = iris.values[:, 0:4]
y = iris.values[:, 4]

# Make test and train set
train_X, test_X, train_y, test_y = train_test_split(X, y, train_size=0.5, random_state=0)

################################
# Evaluate Logistic Regression
################################
lr = LogisticRegressionCV()
lr.fit(train_X, train_y)
pred_y = lr.predict(test_X)
print("Test fraction correct (LR-Accuracy) = {:.2f}".format(lr.score(test_X, test_y)))



################################
# Evaluate Keras Neural Network
################################

# Make ONE-HOT
def one_hot_encode_object_array(arr):
    '''One hot encode a numpy array of objects (e.g. strings)'''
    uniques, ids = np.unique(arr, return_inverse=True)
    return np_utils.to_categorical(ids, len(uniques))


train_y_ohe = one_hot_encode_object_array(train_y)
test_y_ohe = one_hot_encode_object_array(test_y)

model = Sequential()
model.add(Dense(16, input_shape=(4,)))
model.add(Activation('sigmoid'))
model.add(Dense(3))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')

# Actual modelling
model.fit(train_X, train_y_ohe, verbose=0, batch_size=1)
score, accuracy = model.evaluate(test_X, test_y_ohe, batch_size=16, verbose=0)
print("Test fraction correct (NN-Score) = {:.2f}".format(score))
print("Test fraction correct (NN-Accuracy) = {:.2f}".format(accuracy))

I'm using this version of Keras

In [2]: keras.__version__
Out[2]: '1.0.1'

The result shows:

Test fraction correct (LR-Accuracy) = 0.83
Test fraction correct (NN-Score) = 0.75
Test fraction correct (NN-Accuracy) = 0.60

According to that post , the accuracy of Keras should be 0.99. What went wrong?

Answer 1

Your neural network is quite simple. Try creating Deep neural network by adding more neurons and layers into it. Also, it's important to scale your features. Try glorot_uniform initializer. Last but not least, increase epoch and see if loss is decreasing with each epoch.

So here you go:

model = Sequential()
model.add(Dense(input_dim=4, output_dim=512, init='glorot_uniform'))
model.add(PReLU(input_shape=(512,)))
model.add(BatchNormalization((512,)))
model.add(Dropout(0.5))

model.add(Dense(input_dim=512, output_dim=512, init='glorot_uniform'))
model.add(PReLU(input_shape=(512,)))
model.add(BatchNormalization((512,)))
model.add(Dropout(0.5))

model.add(Dense(input_dim=512, output_dim=512, init='glorot_uniform'))
model.add(PReLU(input_shape=(512,)))
model.add(BatchNormalization((512,)))
model.add(Dropout(0.5))

model.add(Dense(input_dim=512, output_dim=512, init='glorot_uniform'))
model.add(PReLU(input_shape=(512,)))
model.add(BatchNormalization((512,)))
model.add(Dropout(0.5))

model.add(Dense(input_dim=512, output_dim=512, init='glorot_uniform'))
model.add(PReLU(input_shape=(512,)))
model.add(BatchNormalization((512,)))
model.add(Dropout(0.5))

model.add(Dense(input_dim=512, output_dim=3, init='glorot_uniform'))
model.add(Activation('softmax'))

This reaches around 0.97 in 120th epoch

Answer 2

The default number of epochs was reduced from 100 in Keras version 0 to 10 in Keras version 1, just released this month (April 2016). Try:

model.fit(train_X, train_y_ohe, verbose=0, batch_size=1, nb_epoch=100)