Tensorflow神經網絡之多層感知機

多層感知機簡介

多層感知機模型

這裏定義含有兩個隱含層的模型，隱含層輸出均爲256個節點，輸入784（MNIST數據集圖片大小28*28），輸出10。

激活函數

比較常用的是 ReLU：relu(x)=max(x,0)，本例中沒有加激活函數。

softmax（同前面的logistic迴歸）

損失函數：交叉熵

Tensorflow實現多層感知機

from __future__ import print_function

import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np

導入數據集

# Import MNIST data
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("./data/", one_hot=False)

Extracting ./data/train-images-idx3-ubyte.gz
Extracting ./data/train-labels-idx1-ubyte.gz
Extracting ./data/t10k-images-idx3-ubyte.gz
Extracting ./data/t10k-labels-idx1-ubyte.gz

設置參數

# Parameters
learning_rate = 0.1
num_steps = 1000
batch_size = 128
display_step = 100

# Network Parameters
n_hidden_1 = 256 # 1st layer number of neurons
n_hidden_2 = 256 # 2nd layer number of neurons
num_input = 784 # MNIST data input (img shape: 28*28)
num_classes = 10 # MNIST total classes (0-9 digits)

定義多層感知機模型

# Define the neural network
def neural_net(x_dict):
    # TF Estimator input is a dict, in case of multiple inputs
    x = x_dict['images']
    # Hidden fully connected layer with 256 neurons
    layer_1 = tf.layers.dense(x, n_hidden_1)
    # Hidden fully connected layer with 256 neurons
    layer_2 = tf.layers.dense(layer_1, n_hidden_2)
    # Output fully connected layer with a neuron for each class
    out_layer = tf.layers.dense(layer_2, num_classes)
    return out_layer

補充：dense [^2]

dense(
   inputs,
   units,
   activation=None,
   use_bias=True,
   kernel_initializer=None,
   bias_initializer=tf.zeros_initializer(),
   kernel_regularizer=None,
   bias_regularizer=None,
   activity_regularizer=None,
   kernel_constraint=None,
   bias_constraint=None,
   trainable=True,
   name=None,
   reuse=None
)

參數說明如下：

• inputs：必需，即需要進行操作的輸入數據。
• units：必須，即神經元的數量。
• activation：可選，默認爲 None，如果爲 None 則是線性激>活。
• use_bias：可選，默認爲 True，是否使用偏置。
• kernel_initializer：可選，默認爲 None，即權重的初始化>方法，如果爲 None，則使用默認的 Xavier 初始化方法。
• bias_initializer：可選，默認爲零值初始化，即偏置的初始>化方法。
• kernel_regularizer：可選，默認爲 None，施加在權重上的>正則項。
• bias_regularizer：可選，默認爲 None，施加在偏置上的正>則項。
• activity_regularizer：可選，默認爲 None，施加在輸出上>的正則項。
• kernel_constraint，可選，默認爲 None，施加在權重上的>約束項。
• bias_constraint，可選，默認爲 None，施加在偏置上的約束>項。
• trainable：可選，默認爲 True，布爾類型，如果爲 True，>則將變量添加到 GraphKeys.TRAINABLE_VARIABLES 中。
• name：可選，默認爲 None，卷積層的名稱。
• reuse：可選，默認爲 None，布爾類型，如果爲 True，那麼>如果 name 相同時，會重複利用。
• 返回值： 全連接網絡處理後的 Tensor。

上面的代碼中，第三個參數爲空，所以這裏採用線性激活

定義模型函數

# Define the model function (following TF Estimator Template)
def model_fn(features, labels, mode):

    # Build the neural network
    logits = neural_net(features)

    # Predictions
    pred_classes = tf.argmax(logits, axis=1)
    pred_probas = tf.nn.softmax(logits)

    # If prediction mode, early return
    if mode == tf.estimator.ModeKeys.PREDICT:
        return tf.estimator.EstimatorSpec(mode, predictions=pred_classes)

    # Define loss and optimizer
    loss_op = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(
        logits=logits, labels=tf.cast(labels, dtype=tf.int32)))
    optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
    train_op = optimizer.minimize(loss_op, global_step=tf.train.get_global_step())

    # Evaluate the accuracy of the model
    acc_op = tf.metrics.accuracy(labels=labels, predictions=pred_classes)

    # TF Estimators requires to return a EstimatorSpec, that specify
    # the different ops for training, evaluating, ...
    estim_specs = tf.estimator.EstimatorSpec(
      mode=mode,
      predictions=pred_classes,
      loss=loss_op,
      train_op=train_op,
      eval_metric_ops={'accuracy': acc_op})

    return estim_specs

構建評估器

# Build the Estimator
model = tf.estimator.Estimator(model_fn)

INFO:tensorflow:Using default config.
WARNING:tensorflow:Using temporary folder as model directory: C:\Users\xywang\AppData\Local\Temp\tmp995gddib
INFO:tensorflow:Using config: {'_model_dir': 'C:\\Users\\xywang\\AppData\\Local\\Temp\\tmp995gddib', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': None, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x0000024BBEE7A1D0>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}

補充：tf.estimator
tf.estimator就是已經預定義好的模型，自帶train，evaluate，predict。
其編程範式爲：

• 定義算法模型，比如多層感知機，CNN；
• 定義模型函數（model_fn），包括構建graph，定義損失函數、優化器，估計準確率等，返回結果分訓練和測試兩種情況；
• 構建評估器；

model = tf.estimator.Estimator(model_fn)

• 用 tf.estimator.inputs.numpy_input_fn 把 input_fn 傳入 model，就可調用 model.train, model.evaluate, model.predict 了。

input_fn = tf.estimator.inputs.numpy_input_fn()
model.train(input_fn)
model.evaluate(input_fn)
model.predict(input_fn)

Estimator 是一種更高層次的封裝，它把一些基本算法的算法>模型和模型函數預定義好，你只需要傳入參數即可。

input_fn [^1]

一般來講，input_fn方法做兩件事：

1.數據預處理，如洗髒數據，歸整數據等。沒有就空着。

2.返回feature_cols, labels。

• feature_cols：一個dict，key爲feature名，value>爲feature值。
• lables: 對應的分類標籤。

可以將多種對象轉換爲tensorflow對象，常見的爲將Numpy>轉tensorflow對象。比如：

import tensorflow as tf
import numpy as np
#numpy input_fn.
x_data =[{"feature1": 2, "features2":6},
         {"feature1": 1, "features2":5},
         {"feature1": 4, "features2":8}]
y_data = [2,3,4]
my_input_fn = tf.estimator.inputs.numpy_input_fn(
    x={"x": np.array(x_data)},
    y=np.array(y_data),
    shuffle = True)
#得到的是一個名爲my_input_fn的function對象。

訓練模型

# Define the input function for training
input_fn = tf.estimator.inputs.numpy_input_fn(
    x={'images': mnist.train.images}, y=mnist.train.labels,
    batch_size=batch_size, num_epochs=None, shuffle=True)

#Train the Model
model.train(input_fn, steps=num_steps)

INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Saving checkpoints for 1 into C:\Users\xywang\AppData\Local\Temp\tmp995gddib\model.ckpt.
INFO:tensorflow:loss = 2.4631722, step = 1
INFO:tensorflow:global_step/sec: 248.501
INFO:tensorflow:loss = 0.42414927, step = 101 (0.388 sec)
INFO:tensorflow:global_step/sec: 301.862
INFO:tensorflow:loss = 0.48539022, step = 201 (0.331 sec)
INFO:tensorflow:global_step/sec: 310.3
INFO:tensorflow:loss = 0.2779913, step = 301 (0.323 sec)
INFO:tensorflow:global_step/sec: 303.697
INFO:tensorflow:loss = 0.20052063, step = 401 (0.329 sec)
INFO:tensorflow:global_step/sec: 322.311
INFO:tensorflow:loss = 0.5092098, step = 501 (0.309 sec)
INFO:tensorflow:global_step/sec: 337.555
INFO:tensorflow:loss = 0.28386787, step = 601 (0.297 sec)
INFO:tensorflow:global_step/sec: 322.309
INFO:tensorflow:loss = 0.36957514, step = 701 (0.309 sec)
INFO:tensorflow:global_step/sec: 334.17
INFO:tensorflow:loss = 0.28504127, step = 801 (0.300 sec)
INFO:tensorflow:global_step/sec: 319.222
INFO:tensorflow:loss = 0.37339848, step = 901 (0.312 sec)
INFO:tensorflow:Saving checkpoints for 1000 into C:\Users\xywang\AppData\Local\Temp\tmp995gddib\model.ckpt.
INFO:tensorflow:Loss for final step: 0.2043538.

tensorflow.python.estimator.estimator.Estimator at 0x24bbee72160

模型評估

# Evaluate the Model
# Define the input function for evaluating
input_fn = tf.estimator.inputs.numpy_input_fn(
    x={'images': mnist.test.images}, y=mnist.test.labels,
    batch_size=batch_size, shuffle=False)
# Use the Estimator 'evaluate' method
model.evaluate(input_fn)

INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Starting evaluation at 2018-04-11-08:04:38
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from C:\Users\xywang\AppData\Local\Temp\tmp995gddib\model.ckpt-1000
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Finished evaluation at 2018-04-11-08:04:38
INFO:tensorflow:Saving dict for global step 1000: accuracy = 0.9149, global_step = 1000, loss = 0.29386005

{'accuracy': 0.9149, 'global_step': 1000, 'loss': 0.29386005}

測試

# Predict single images
n_images = 1
# Get images from test set
test_images = mnist.test.images[:n_images]
# Prepare the input data
input_fn = tf.estimator.inputs.numpy_input_fn(
    x={'images': test_images}, shuffle=False)
# Use the model to predict the images class
preds = list(model.predict(input_fn))

# Display
for i in range(n_images):
    plt.imshow(np.reshape(test_images[i], [28, 28]), cmap='gray')
    plt.show()
    print("Model prediction:", preds[i])

INFO:tensorflow:Calling model_fn.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Restoring parameters from C:\Users\xywang\AppData\Local\Temp\tmp995gddib\model.ckpt-1000
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.

Model prediction: 7

參考

[1] 在tf.estimator中構建inpu_fn解讀

[2] TensorFlow layers模塊用法

---