In [1]:
#!/anaconda3/envs/py36/bin/python3
import numpy as np
import pandas as pd
import tensorflow as tf
import warnings
warnings.filterwarnings("ignore")
In [2]:
iris = pd.read_csv("./datas/iris.csv")
In [3]:
iris_target_names = np.unique(iris["species"].values)
In [4]:
for i, n in enumerate(iris_target_names):
iris.replace(to_replace=iris_target_names[i], value=i, inplace=True)
In [5]:
from sklearn.model_selection import train_test_split
train_set, test_set = train_test_split(iris, test_size=0.2, random_state=42)
train_labels = train_set["species"].values
test_labels = test_set["species"].values
In [6]:
import matplotlib.pyplot as plt
%matplotlib inline
train_set.plot(kind="scatter", x="sepal_length", y="sepal_width", alpha=0.4, figsize=(10, 8), c="species", cmap=plt.get_cmap("jet"), colorbar=True, sharex=False)
plt.show()
In [7]:
train_set.hist(bins=50, figsize=(20, 15))
plt.show()
In [8]:
from sklearn.preprocessing import OneHotEncoder, StandardScaler, LabelEncoder
from sklearn.impute import SimpleImputer
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.pipeline import Pipeline, FeatureUnion
class DataFrameSelector(BaseEstimator, TransformerMixin):
def __init__(self, lists):
self.lists = lists
def fit(self, X, y=None):
return self
def transform(self, X):
return X.iloc[:, self.lists].values
# class LabelEncoders(BaseEstimator, TransformerMixin):
# def __init__(self):
# self.encoder = LabelEncoder()
# def fit(self, X, y=None):
# self.encoder.fit(X)
# return self
# def transform(self, X, y=None):
# return self.encoder.transform(X)
In [9]:
n_list = [0, 1, 2, 3]
e_list = [4]
In [10]:
train_set["species"] = train_set["species"].astype(str)
In [11]:
num_pipeline = Pipeline([
["selector", DataFrameSelector(lists=n_list)],
["imputer" , SimpleImputer(strategy="median")],
["scaler" , StandardScaler()]
])
encoding_pipeline = Pipeline([
["selector", DataFrameSelector(lists=e_list)],
["encoder", OneHotEncoder(sparse=False, categories="auto")]
])
full_pipeline = FeatureUnion(transformer_list=[
["nums", num_pipeline],
["encoding", encoding_pipeline]
])
In [12]:
scaled_train = full_pipeline.fit_transform(train_set)
scaled_test = full_pipeline.fit_transform(test_set)
In [13]:
x_train, y_train = scaled_train[:, :4], scaled_train[:, 4:]
x_test, y_test = scaled_test[:, :4], scaled_test[:, 4:]
In [14]:
global_step = tf.Variable(0, trainable=False, name="global_step")
In [15]:
X = tf.placeholder(tf.float32)
Y = tf.placeholder(tf.float32)
In [16]:
# 4, 1
W1 = tf.Variable(tf.random_normal([4, 10], mean=0, stddev=1))
W2 = tf.Variable(tf.random_normal([10, 100], mean=0, stddev=1))
W3 = tf.Variable(tf.random_normal([100, 3], mean=0, stddev=1))
b1 = tf.zeros([10])
b2 = tf.zeros([100])
b3 = tf.zeros([3])
In [17]:
# Layer 1
with tf.name_scope("Layer1"):
L1 = tf.add(tf.matmul(X, W1), b1)
L1 = tf.nn.sigmoid(L1)
with tf.name_scope("Layer2"):
L2 = tf.add(tf.matmul(L1, W2), b2)
L2 = tf.nn.sigmoid(L2)
with tf.name_scope("Layer3"):
model = tf.add(tf.matmul(L2, W3), b3)
model = tf.nn.sigmoid(model)
In [18]:
with tf.name_scope("optiimizer"):
cost = tf.reduce_mean(-tf.reduce_sum(Y*tf.log(model) + (1-Y)*tf.log(1-model)))
optimizer = tf.train.AdadeltaOptimizer(learning_rate=0.01)
train_op = optimizer.minimize(cost, global_step=global_step)
tf.summary.scalar("cost", cost)
tf.summary.histogram("Weight1", W1)
tf.summary.histogram("Weight2", W2)
tf.summary.histogram("Weight3", W3)
tf.summary.histogram("bias1", b1)
tf.summary.histogram("bias2", b2)
tf.summary.histogram("bias3", b3)
In [19]:
sess = tf.Session()
saver = tf.train.Saver(tf.global_variables())
In [20]:
ckpt = tf.train.get_checkpoint_state("./model/iris")
if ckpt and tf.train.checkpoint_exists(ckpt.model_checkpoint_path):
saver.restore(sess, ckpt.model_checkpoint_path)
else:
sess.run(tf.global_variables_initializer())
In [21]:
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter("./logs/iris", sess.graph)
In [22]:
for step in range(5000):
sess.run(train_op, feed_dict={X:x_train, Y:y_train})
if (step+1) % 200 == 0:
print("step: {}, cost: {:.5f}".\
format(sess.run(global_step),
sess.run(cost, feed_dict={X:x_train, Y:y_train})))
summary = sess.run(merged, feed_dict={X:x_train, Y:y_train})
writer.add_summary(summary, global_step=sess.run(global_step))
saver.save(sess, "./model/iris/dnn.ckpt", global_step=global_step)
Out[22]:
In [23]:
# print("{}".format(sess.run(model, feed_dict={X:x_scaled_test, Y:y_test})))
prediction = tf.argmax(model, 1)
target = tf.argmax(Y, 1)
is_correct = tf.equal(prediction, target)
accuracy = tf.reduce_mean(tf.cast(is_correct, tf.float32))
print("====================================")
print("train_prediction: \n{}".format(sess.run(prediction, feed_dict={X: x_train})))
print("train_target: \n{}\n".format(sess.run(target, feed_dict={Y:y_train})))
print("\naccuracy: \n{:.3f}%".format(sess.run(accuracy*100,
feed_dict={X: x_train, Y:y_train})))
print("\n====================================")
print("test_prediction: \n{}".format(sess.run(prediction, feed_dict={X: x_test})))
print("test_target: \n{}".format(sess.run(target, feed_dict={Y:y_test})))
print("\naccuracy: \n{:.3f}%".format(sess.run(accuracy*100,
feed_dict={X: x_test, Y:y_test})))
In [24]:
# tensorboard --logdir=./logs/iris
In [25]:
from IPython.core.display import display, HTML
display(HTML("<style> .container{width:100% !important;}</style>"))
'Deep_Learning' 카테고리의 다른 글
| 09.mnist_01_minibatch (0) | 2018.12.09 |
|---|---|
| 08.tensorboard03_example (0) | 2018.12.09 |
| 06.tensorboard01_example (0) | 2018.12.09 |
| 05.deep_neural_net_Costfun2 (0) | 2018.12.09 |
| 04.deep_neural_net_Costfun1 (0) | 2018.12.09 |