from matplotlib import pyplot as plt
"""
Experiment Description
----------------------
Return display the initial and final weights of the network
"""
P = 200
Q = 100
learning_rate = 0.05
t_max = 100 #total epochs
trainX, trainY, testX, testY = read_file("xi(1).csv", "tau(1).csv", P, Q)
model = Model(input_size=len(trainX[0]))
model.add_layer(states=2, activation='tanh', fixed_weights=False)
model.add_layer(states=1, activation=None, fixed_weights=1)
# model.display()
# print(trainX.shape)
initial_weights, final_weights = model.train(trainX,
trainY,
testX,
testY,
ephochs=t_max,
learning_rate=learning_rate,
return_weights=True,
verbose=False)
R = 50
P = 200
Q = 100
learning_rate = 0.05
t_max = 100 #total epochs
trainX, trainY, testX, testY = read_file("xi(1).csv", "tau(1).csv", P, Q)
Es = []
E_tests = []
for i in range(0, R):
print("Run " + str(i + 1) + " of " + str(R))
model = Model(input_size=len(trainX[0]))
model.add_layer(states=2, activation='tanh', fixed_weights=False)
model.add_layer(states=1, activation=None, fixed_weights=1)
# model.display()
# print(trainX.shape)
E, E_test = model.train(trainX,
trainY,
testX,
testY,
ephochs=t_max,
learning_rate=learning_rate)
Es.append(E)
E_tests.append(E_test)
from Network import Model
from PSO import optimize
from InputParam import InputParam
import numpy as np
import time
data = np.loadtxt('./data/data.csv', delimiter=',', skiprows=2)[:, 1:7]
model = Model(stride=30)
param = InputParam()
start = time.clock()
# 调用optimize
optimize(20, data, 2, model, param)
end = time.clock()
from Network import Model
import numpy as np
'''原Interface'''
data = np.loadtxt('./data/data.csv', delimiter=',', skiprows=2)[:, 1: 7]
# 创建 Model， stride就是代表历史长度，通过更改model.stride就可以实现对模型的选择
# 例如: model.stride = n（n的取值范围5,10,15,20,25,30,35,40,45,50）
model = Model(stride=30)
#全连接神经网络 模型预测
# data_x是输入，输入的尺寸是（n, 6)，n是数据的个数
model.stride = 25
data_x = data[-25:, :]
# data_y是输出的预测的结果，输出的矩阵的形状是(n, 4), n是数据的个数
data_y = model.predict_LSTM(data_x)
print(data_y)
# LSTM 模型预测
# data_x是输入，输入的矩阵形状是（n,stride*6) stride是步长，n是数据的个数
#data_x = [[1,2,3,4,5,6],[1,2,3,4,5,6],[1,2,3,4,5,6],[1,2,3,4,5,6],[1,2,3,4,5,6]]
# data_y是输出的预测的结果，输出的矩阵的形状是(4,n), n是数据的个数
#model.stride = 5
data_y = model.predict_Net(data_x)
print(type(data_y[0]))
############################################
T_len, H2_len, CH4_len, CO_len = 5,10,5,20
from Network import Model
import numpy as np
'''原Interface'''
data = np.loadtxt('./data/data.csv', delimiter=',', skiprows=2)[:, 1:7]
# 创建 Model， stride就是代表历史长度，通过更改model.stride就可以实现对模型的选择
# 例如: model.stride = n（n的取值范围5,10,15,20,25,30,35,40,45,50）
model = Model(stride=30)
#全连接神经网络 模型预测
# data_x是输入，输入的尺寸是（n, 6)，n是数据的个数
model.stride = 25
data_x = data[-25:, :]
# data_y是输出的预测的结果，输出的矩阵的形状是(n, 4), n是数据的个数
data_y = model.predict_Net(data_x)
print(data_y)
# LSTM 模型预测
# data_x是输入，输入的矩阵形状是（n,stride*6) stride是步长，n是数据的个数
data_x = [[1, 2, 3, 4, 5, 6], [1, 2, 3, 4, 5, 6], [1, 2, 3, 4, 5, 6],
[1, 2, 3, 4, 5, 6], [1, 2, 3, 4, 5, 6]]
# data_y是输出的预测的结果，输出的矩阵的形状是(4,n), n是数据的个数
model.stride = 5
data_y = model.test_future_LSTM()
print(data_y)
from Network import Model
from PSO import optimize
from InputParam import InputParam
import numpy as np
import time
data = np.loadtxt('./data/data.csv', delimiter=',', skiprows=2)[:, 1:7]
model = Model(stride=30)
param = InputParam()
best_x = [[28.12421607, 14.41792793], [74.20071455, 9.92461491],
[58.43944117, 11.0825983], [69.17058085, 10.75863371],
[36.35079316, 11.28547075], [57.59862793, 10.92375781],
[27.67301199, 14.30238826], [75.87876121, 13.22701089],
[100.0, 16.10638031], [100.0, 13.78599625]]
best_y = 26.45409999049893
# 调用optimize
#best_x, best_y = optimize(20, data, 2, model, param)
print('best_x = ', best_x)
print('best_y = ', best_y)
model.stride = 10
predict_y = model.predict_Net(best_x)
print(predict_y)
from tensorflow.keras.datasets import mnist
from tensorflow.keras.utils import to_categorical
if __name__ == "__main__":
# Load data
(x_train, y_train), (x_test, y_test) = mnist.load_data()
# Flatten input 28 * 28 matrix into a 784 vector
x_train = x_train.reshape(len(x_train), 28 * 28)
x_test = x_test.reshape(len(x_test), 28 * 28)
# Make y into one-hot encoding
y_train = to_categorical(y_train, 10)
y_test = to_categorical(y_test, 10)
# Build model
model = Model()
model.add_layer(in_dim=28 * 28, out_dim=500, activation="tanh")
model.add_layer(out_dim=500, activation="sigmoid")
model.add_layer(out_dim=10, activation="softmax")
# Compile model
model.compile(loss="cross entropy")
# Train model
model.fit(x_train, y_train, batch_size=256, epochs=5)
# Predict and evaluate
loss, accuracy = model.evaluate(x_test, y_test)
print("Total loss for prediction: {}, accuracy: {}".format(loss, accuracy))