本文接上一篇博文:
一、程序代碼
程序主要實現上篇文章中所提到的隨機噪聲擬合高斯分佈的過程,話不多說,直接上代碼:
#引入必要的包
import argparse
import numpy as np
from scipy.stats import norm
import tensorflow as tf
import matplotlib.pyplot as plt
import seaborn as sns
sns.set(color_codes=True)
#設置種子,用於隨機初始化
seed = 42
np.random.seed(seed)
tf.set_random_seed(seed)
#定義真實的數據分佈,這裏爲高斯分佈
class DataDistribution(object):
def __init__(self):
#高斯分佈參數
#均值爲4
self.mu = 4
#標準差爲0.5
self.sigma = 0.5
def sample(self, N):
samples = np.random.normal(self.mu, self.sigma, N)
samples.sort()
return samples
#隨機初始化一個分佈,做爲G網絡的輸入
class GeneratorDistribution(object):
def __init__(self, range):
self.range = range
def sample(self, N):
return np.linspace(-self.range, self.range, N) + \
np.random.random(N) * 0.01
#定義線性運算函數,其中參數output_dim=h_dim*2=8
def linear(input, output_dim, scope=None, stddev=1.0):
#定義一個隨機初始化
norm = tf.random_normal_initializer(stddev=stddev)
#b初始化爲0
const = tf.constant_initializer(0.0)
with tf.variable_scope(scope or 'linear'):
#聲明w的shape,輸入爲(12,1)*w,故w爲(1,8),w的初始化方式爲高斯初始化
w = tf.get_variable('w', [input.get_shape()[1], output_dim], initializer=norm)
#b初始化爲常量
b = tf.get_variable('b', [output_dim], initializer=const)
#執行線性運算
return tf.matmul(input, w) + b
#
def generator(input, h_dim):
h0 = tf.nn.softplus(linear(input, h_dim, 'g0'))
h1 = linear(h0, 1, 'g1')
return h1
#初始化w和b的函數,其中h0,h1,h2,h3爲層,將mlp_hidden_size=4傳給h_dim
def discriminator(input, h_dim):
#linear 控制w和b的初始化,這裏linear函數的第二個參數爲4*2=8
#第一層
h0 = tf.tanh(linear(input, h_dim * 2, 'd0'))
#第二層輸出,隱藏層神經元個數還是爲8
h1 = tf.tanh(linear(h0, h_dim * 2, 'd1'))
#h2爲第三層輸出值
h2 = tf.tanh(linear(h1, h_dim * 2, scope='d2'))
#最終的輸出值
h3 = tf.sigmoid(linear(h2, 1, scope='d3'))
return h3
#優化器 採用學習率衰減的方法
def optimizer(loss, var_list, initial_learning_rate):
decay = 0.95
num_decay_steps = 150
batch = tf.Variable(0)
#調用學習率衰減的函數
learning_rate = tf.train.exponential_decay(
initial_learning_rate,
batch,
num_decay_steps,
decay,
staircase=True
)
#梯度下降求解
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(
loss,
global_step=batch,
var_list=var_list
)
#返回
return optimizer
#構造模型
class GAN(object):
def __init__(self, data, gen, num_steps, batch_size, log_every):
self.data = data
self.gen = gen
self.num_steps = num_steps
self.batch_size = batch_size
self.log_every = log_every
#隱藏層神經元個數
self.mlp_hidden_size = 4
#學習率
self.learning_rate = 0.03
#通過placeholder格式來創造模型
self._create_model()
def _create_model(self):
#創建一個名叫D_pre的域,先構造一個D_pre網絡,用來訓練出真正D網絡初始化網絡所需要的參數
with tf.variable_scope('D_pre'):
#輸入的shape爲(12,1),一個batch一個batch的訓練,
#每個batch的大小爲12,要訓練的數據爲1維的點
self.pre_input = tf.placeholder(tf.float32, shape=(self.batch_size, 1))
self.pre_labels = tf.placeholder(tf.float32, shape=(self.batch_size, 1))
#調用discriminator來初始化w和b參數,其中self.mlp_hidden_size=4,爲discriminator函數的第二個參數
D_pre = discriminator(self.pre_input, self.mlp_hidden_size)
#預測值和label之間的差異
self.pre_loss = tf.reduce_mean(tf.square(D_pre - self.pre_labels))
#定義優化器求解
self.pre_opt = optimizer(self.pre_loss, None, self.learning_rate)
# This defines the generator network - it takes samples from a noise
# distribution as input, and passes them through an MLP.
#真正的G網絡
with tf.variable_scope('Gen'):
self.z = tf.placeholder(tf.float32, shape=(self.batch_size, 1))
#生產網絡只有兩層
self.G = generator(self.z, self.mlp_hidden_size)
# The discriminator tries to tell the difference between samples from the
# true data distribution (self.x) and the generated samples (self.z).
#
# Here we create two copies of the discriminator network (that share parameters),
# as you cannot use the same network with different inputs in TensorFlow.
#D網絡
with tf.variable_scope('Disc') as scope:
self.x = tf.placeholder(tf.float32, shape=(self.batch_size, 1))
#構造D1網絡,真實的數據
self.D1 = discriminator(self.x, self.mlp_hidden_size)
#重新使用一下變量,不用重新定義
scope.reuse_variables()
#D2,生成的數據
self.D2 = discriminator(self.G, self.mlp_hidden_size)
# Define the loss for discriminator and generator networks (see the original
# paper for details), and create optimizers for both
#定義判別網絡損失函數
self.loss_d = tf.reduce_mean(-tf.log(self.D1) - tf.log(1 - self.D2))
#定義生成網絡損失函數
self.loss_g = tf.reduce_mean(-tf.log(self.D2))
self.d_pre_params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope='D_pre')
self.d_params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope='Disc')
self.g_params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope='Gan')
#優化,得到兩組參數
self.opt_d = optimizer(self.loss_d, self.d_params, self.learning_rate)
self.opt_g = optimizer(self.loss_g, self.g_params, self.learning_rate)
def train(self):
with tf.Session() as session:
tf.global_variables_initializer().run()
# pretraining discriminator
#先訓練D_pre網絡
num_pretrain_steps = 1000
for step in range(num_pretrain_steps):
#隨機生成數據
d = (np.random.random(self.batch_size) - 0.5) * 10.0
labels = norm.pdf(d, loc=self.data.mu, scale=self.data.sigma)
pretrain_loss, _ = session.run([self.pre_loss, self.pre_opt], {
self.pre_input: np.reshape(d, (self.batch_size, 1)),
self.pre_labels: np.reshape(labels, (self.batch_size, 1))
})
#拿出預訓練好的數據
self.weightsD = session.run(self.d_pre_params)
# copy weights from pre-training over to new D network
for i, v in enumerate(self.d_params):
session.run(v.assign(self.weightsD[i]))
#訓練真正的網絡
for step in range(self.num_steps):
# update discriminator
x = self.data.sample(self.batch_size)
#z是一個隨機生成的噪音
z = self.gen.sample(self.batch_size)
#優化判別網絡
loss_d, _ = session.run([self.loss_d, self.opt_d], {
self.x: np.reshape(x, (self.batch_size, 1)),
self.z: np.reshape(z, (self.batch_size, 1))
})
# update generator
#隨機初始化
z = self.gen.sample(self.batch_size)
#迭代優化
loss_g, _ = session.run([self.loss_g, self.opt_g], {
self.z: np.reshape(z, (self.batch_size, 1))
})
#打印
if step % self.log_every == 0:
print('{}: {}\t{}'.format(step, loss_d, loss_g))
#畫圖
if step % 100 == 0 or step==0 or step == self.num_steps -1 :
self._plot_distributions(session)
def _samples(self, session, num_points=10000, num_bins=100):
xs = np.linspace(-self.gen.range, self.gen.range, num_points)
bins = np.linspace(-self.gen.range, self.gen.range, num_bins)
# data distribution
d = self.data.sample(num_points)
pd, _ = np.histogram(d, bins=bins, density=True)
# generated samples
zs = np.linspace(-self.gen.range, self.gen.range, num_points)
g = np.zeros((num_points, 1))
for i in range(num_points // self.batch_size):
g[self.batch_size * i:self.batch_size * (i + 1)] = session.run(self.G, {
self.z: np.reshape(
zs[self.batch_size * i:self.batch_size * (i + 1)],
(self.batch_size, 1)
)
})
pg, _ = np.histogram(g, bins=bins, density=True)
return pd, pg
def _plot_distributions(self, session):
pd, pg = self._samples(session)
p_x = np.linspace(-self.gen.range, self.gen.range, len(pd))
f, ax = plt.subplots(1)
ax.set_ylim(0, 1)
plt.plot(p_x, pd, label='real data')
plt.plot(p_x, pg, label='generated data')
plt.title('1D Generative Adversarial Network')
plt.xlabel('Data values')
plt.ylabel('Probability density')
plt.legend()
plt.show()
def main(args):
model = GAN(
#定義真實數據的分佈
DataDistribution(),
#創造一些噪音點,用來傳入G函數
GeneratorDistribution(range=8),
#迭代次數
args.num_steps,
#一次迭代12個點的數據
args.batch_size,
#隔多少次打印當前loss
args.log_every,
)
model.train()
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--num-steps', type=int, default=3000,
help='the number of training steps to take')
parser.add_argument('--batch-size', type=int, default=12,
help='the batch size')
parser.add_argument('--log-every', type=int, default=10,
help='print loss after this many steps')
return parser.parse_args()
if __name__ == '__main__':
main(parse_args())
二、程序運行結果
1、程序運行初始狀態
其中左邊爲隨機初始化的數據,右邊爲真實的呈高斯分佈的數據。
2、程序迭代運行1200次後的狀態
這裏不知道爲什麼原因,程序沒有正常的擬合真實的數據,將迭代次數增加之後,程序也沒有太大的變化,D和G網絡的兩個Loss的變化都很小,這裏還望大家幫忙找一找原因。可能和GAN網絡容易訓練跑偏的一些原因有關。