memory.py

#  #################################################################
#  This file contains memory operation including encoding and decoding operations.
#
# version 1.0 -- January 2018. Written by Liang Huang (lianghuang AT zjut.edu.cn)
#  #################################################################

from __future__ import print_function
import tensorflow as tf
import numpy as np


# DNN network for memory
class MemoryDNN:
    def __init__(
        self,
        net,
        learning_rate = 0.01,
        training_interval=10, 
        batch_size=100, 
        memory_size=1000,
        output_graph=False
    ):
        # net: [n_input, n_hidden_1st, n_hidded_2ed, n_output]
        assert(len(net) is 4) # only 4-layer DNN

        self.net = net
        self.training_interval = training_interval # learn every #training_interval
        self.lr = learning_rate
        self.batch_size = batch_size
        self.memory_size = memory_size
        
        # store all binary actions
        self.enumerate_actions = []

        # stored # memory entry
        self.memory_counter = 1

        # store training cost
        self.cost_his = []

        # reset graph 
        tf.reset_default_graph()

        # initialize zero memory [h, m]
        self.memory = np.zeros((self.memory_size, self.net[0]+ self.net[-1]))

        # construct memory network
        self._build_net()

        self.sess = tf.Session()

        # for tensorboard
        if output_graph:
            # $ tensorboard --logdir=logs
            # tf.train.SummaryWriter soon be deprecated, use following
            tf.summary.FileWriter("logs/", self.sess.graph)

        self.sess.run(tf.global_variables_initializer())


    def _build_net(self):
        def build_layers(h, c_names, net, w_initializer, b_initializer):
            with tf.variable_scope('l1'):
                w1 = tf.get_variable('w1', [net[0], net[1]], initializer=w_initializer, collections=c_names)
                b1 = tf.get_variable('b1', [1, self.net[1]], initializer=b_initializer, collections=c_names)
                l1 = tf.nn.relu(tf.matmul(h, w1) + b1)

            with tf.variable_scope('l2'):
                w2 = tf.get_variable('w2', [net[1], net[2]], initializer=w_initializer, collections=c_names)
                b2 = tf.get_variable('b2', [1, net[2]], initializer=b_initializer, collections=c_names)
                l2 = tf.nn.relu(tf.matmul(l1, w2) + b2)

            with tf.variable_scope('M'):
                w3 = tf.get_variable('w3', [net[2], net[3]], initializer=w_initializer, collections=c_names)
                b3 = tf.get_variable('b3', [1, net[3]], initializer=b_initializer, collections=c_names)
                out = tf.matmul(l2, w3) + b3

            return out

        # ------------------ build memory_net ------------------
        self.h = tf.placeholder(tf.float32, [None, self.net[0]], name='h')  # input
        self.m = tf.placeholder(tf.float32, [None, self.net[-1]], name='mode')  # for calculating loss
        self.is_train = tf.placeholder("bool") # train or evaluate

        with tf.variable_scope('memory_net'):
            c_names, w_initializer, b_initializer = \
                ['memory_net_params', tf.GraphKeys.GLOBAL_VARIABLES], \
                tf.random_normal_initializer(0., 1/self.net[0]), tf.constant_initializer(0.1)  # config of layers

            self.m_pred = build_layers(self.h, c_names, self.net, w_initializer, b_initializer)

        with tf.variable_scope('loss'):
            self.loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels = self.m, logits = self.m_pred))

        with tf.variable_scope('train'):
            self._train_op = tf.train.AdamOptimizer(self.lr, 0.09).minimize(self.loss)


    def remember(self, h, m):
        # replace the old memory with new memory
        idx = self.memory_counter % self.memory_size
        self.memory[idx, :] = np.hstack((h,m))

        self.memory_counter += 1

    def encode(self, h, m):
        # encoding the entry
        self.remember(h, m)
        # train the DNN every 10 step
#        if self.memory_counter> self.memory_size / 2 and self.memory_counter % self.training_interval == 0:
        if self.memory_counter % self.training_interval == 0:
            self.learn()

    def learn(self):
        # sample batch memory from all memory
        if self.memory_counter > self.memory_size:
            sample_index = np.random.choice(self.memory_size, size=self.batch_size)
        else:
            sample_index = np.random.choice(self.memory_counter, size=self.batch_size)
        batch_memory = self.memory[sample_index, :]
        
        h_train = batch_memory[:, 0: self.net[0]]
        m_train = batch_memory[:, self.net[0]:]
        
        # print(h_train)
        # print(m_train)

        # train the DNN
        _, self.cost = self.sess.run([self._train_op, self.loss], 
                                         feed_dict={self.h: h_train, self.m: m_train})

        assert(self.cost >0)    
        self.cost_his.append(self.cost)

    def decode(self, h, k = 1, mode = 'OP'):
        # to have batch dimension when feed into tf placeholder
        h = h[np.newaxis, :]

        m_pred = self.sess.run(self.m_pred, feed_dict={self.h: h})

        if mode is 'OP':
            return self.knm(m_pred[0], k)
        elif mode is 'KNN':
            return self.knn(m_pred[0], k)
        else:
            print("The action selection must be 'OP' or 'KNN'")
    
    def knm(self, m, k = 1):
        # return k-nearest-mode
        m_list = []
        
        # generate the first binary offloading decision 
        # note that here 'm' is the output of DNN before the sigmoid activation function, in the field of all real number. 
        # Therefore, we compare it with '0' instead of 0.5 in equation (8). Since, sigmod(0) = 0.5.
        m_list.append(1*(m>0))
        
        if k > 1:
            # generate the remaining K-1 binary offloading decisions with respect to equation (9)
            m_abs = abs(m)
            idx_list = np.argsort(m_abs)[:k-1]
            for i in range(k-1):
                if m[idx_list[i]] >0:
                    # set a positive user to 0
                    m_list.append(1*(m - m[idx_list[i]] > 0))
                else:
                    # set a negtive user to 1
                    m_list.append(1*(m - m[idx_list[i]] >= 0))

        return m_list
    
    def knn(self, m, k = 1):
        # list all 2^N binary offloading actions
        if len(self.enumerate_actions) is 0:
            import itertools
            self.enumerate_actions = np.array(list(map(list, itertools.product([0, 1], repeat=self.net[0]))))

        # the 2-norm
        sqd = ((self.enumerate_actions - m)**2).sum(1)
        idx = np.argsort(sqd)
        return self.enumerate_actions[idx[:k]]
        

    def plot_cost(self):
        import matplotlib.pyplot as plt
        plt.plot(np.arange(len(self.cost_his))*self.training_interval, self.cost_his)
        plt.ylabel('Training Loss')
        plt.xlabel('Time Frames')
        plt.show()