RNN 연습

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

def load_weather(file_path, seq_length):
    weather = np.loadtxt(file_path, delimiter=' ', skiprows=1)
    weather = (weather[:,1:])
    weather = weather.transpose()
    weather = weather.reshape(-1, 1)

    xx, yy = [], []
    for i in range(len(weather)-seq_length):
        xx.append(weather[i:i+seq_length])
        yy.append(weather[i+seq_length])
    return np.float32(xx), np.float32(yy)

# 날씨 데이타# http://www.weather.go.kr/weather/climate/past_table.jspdef rnn_weather():
    batch_size = 1    seq_length = 7    n_classes = 1    hidden_size = 5    output_dim = 1
    xx, yy           = load_weather("Data/weather.txt", seq_length)
    xx_test, yy_test = load_weather("Data/weather_test.txt", seq_length)

    x = tf.placeholder(tf.float32, shape=[None, 7, 1])

    cells = tf.nn.rnn_cell.BasicRNNCell(num_units=hidden_size, activation=tf.tanh)
    output, state_ = tf.nn.dynamic_rnn(cells, x, dtype=tf.float32)

    z = tf.contrib.layers.fully_connected(inputs=output[:, -1],
                                          num_outputs=1,
                                          activation_fn=None)
    loss = tf.reduce_mean((z-yy)**2)

    optimizer = tf.train.AdamOptimizer(0.01)
    train = optimizer.minimize(loss)

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())

        for i in range(5000):
            sess.run(train, {x:xx})
            print(sess.run(loss , {x:xx}))

        p = sess.run(z, {x:xx_test[30:100]})

        plt.plot(p, 'r')
        plt.plot(yy_test[30:100], 'g')
        plt.show()

rnn_weather()