keras 由model.train_on_batch()到model.fit()

于 2022-10-24 21:37:51 发布
阅读量579 收藏 2
点赞数
分类专栏： tensorflow学习
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/k903161661/article/details/127499385
版权

原始代码

'''基本上算是最终版的, 不错的预测, loss<1'''

import csv
from functools import lru_cache
import numpy as np
import matplotlib.pyplot as plt

import keras
from keras import backend as K
from keras.layers import Layer
from keras.layers import Dense
from keras.models import Sequential
from keras.layers import Activation

from pylab import *
from mpl_toolkits.mplot3d import Axes3D

'''设置文件保存路径'''
path = './model_save/4D/Use5/'


'''导入csv文件(返回的数据为ndarry)'''
# 训练集
csv_train = np.loadtxt(open("./train_last.csv","rb"),delimiter=",",skiprows=0) 
print('train.csv导入成功!')
# 测试集
csv_test = np.loadtxt(open("./test_last.csv","rb"),delimiter=",",skiprows=0) 
print('test.csv导入成功!')


'''选取输入向量'''
# 训练集
X_train = csv_train[:,0:3]# [:,np.newaxis]
Y_train = csv_train[:,3]
# 测试集
X_test = csv_test[:,0:3]
Y_test = csv_test[:,3]


# 创建RBF层
class RBFLayer(Layer):
    def __init__(self, units, gamma, **kwargs):
        super(RBFLayer, self).__init__(**kwargs)
        self.units = units
        self.gamma = K.cast_to_floatx(gamma)

    def build(self, input_shape):
        self.mu = self.add_weight(name='mu',
                                  shape=(int(input_shape[1]), self.units),
                                  initializer='uniform',
                                  trainable=True)
        super(RBFLayer, self).build(input_shape)

    def call(self, inputs):
        diff = K.expand_dims(inputs) - self.mu
        l2 = K.sum(K.pow(diff,2), axis=1)
        res = K.exp(-1 * self.gamma * l2)
        return res

    def compute_output_shape(self, input_shape):
        return (input_shape[0], self.units)



'''搭建网络结构'''
'''Glorot 正态分布初始化器，也称为 Xavier 正态分布初始化器。
它从以 0 为中心，标准差为 stddev = sqrt(2 / (fan_in + fan_out)) 的截断正态分布中抽取样本， 
其中 fan_in 是权值张量中的输入单位的数量， fan_out 是权值张量中的输出单位的数量。'''
## keras.initializers.glorot_normal(seed=None) 

model = Sequential()

# 第1层
model.add(Dense(units = 3, input_dim = 3))
model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 8, input_dim = 3))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 16, input_dim = 8))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 32, input_dim = 16))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 64, input_dim = 32))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 128, input_dim = 64))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 256, input_dim = 128))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 512, input_dim = 256))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 1024, input_dim = 512))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 1024, input_dim = 1024))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 512, input_dim = 1024)) 
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 256, input_dim = 512))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 128, input_dim = 256))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 64, input_dim = 128))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 32, input_dim = 64))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 16, input_dim = 32))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 8, input_dim = 16))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 4, input_dim = 8))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 2, input_dim = 4))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 1, input_dim = 64, use_bias=True))
     
# 显示
model.summary()
'''
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_1 (Dense)              (None, 3)                 12
_________________________________________________________________
batch_normalization_1 (Batch (None, 3)                 12
_________________________________________________________________
dense_2 (Dense)              (None, 8)                 32
_________________________________________________________________
batch_normalization_2 (Batch (None, 8)                 32
_________________________________________________________________
activation_1 (Activation)    (None, 8)                 0
_________________________________________________________________
dense_3 (Dense)              (None, 16)                144
_________________________________________________________________
batch_normalization_3 (Batch (None, 16)                64
_________________________________________________________________
activation_2 (Activation)    (None, 16)                0
_________________________________________________________________
dense_4 (Dense)              (None, 32)                544
_________________________________________________________________
batch_normalization_4 (Batch (None, 32)                128
_________________________________________________________________
activation_3 (Activation)    (None, 32)                0
_________________________________________________________________
dense_5 (Dense)              (None, 64)                2112
_________________________________________________________________
batch_normalization_5 (Batch (None, 64)                256
_________________________________________________________________
activation_4 (Activation)    (None, 64)                0
_________________________________________________________________
dense_6 (Dense)              (None, 128)               8320
_________________________________________________________________
batch_normalization_6 (Batch (None, 128)               512
_________________________________________________________________
activation_5 (Activation)    (None, 128)               0
_________________________________________________________________
dense_7 (Dense)              (None, 256)               33024
_________________________________________________________________
batch_normalization_7 (Batch (None, 256)               1024      
_________________________________________________________________
activation_6 (Activation)    (None, 256)               0
_________________________________________________________________
dropout_1 (Dropout)          (None, 256)               0
_________________________________________________________________
dense_8 (Dense)              (None, 512)               131584
_________________________________________________________________
batch_normalization_8 (Batch (None, 512)               2048
_________________________________________________________________
activation_7 (Activation)    (None, 512)               0
_________________________________________________________________
dropout_2 (Dropout)          (None, 512)               0
_________________________________________________________________
dense_9 (Dense)              (None, 1024)              525312
_________________________________________________________________
batch_normalization_9 (Batch (None, 1024)              4096
_________________________________________________________________
activation_8 (Activation)    (None, 1024)              0
_________________________________________________________________
dropout_3 (Dropout)          (None, 1024)              0
_________________________________________________________________
dense_10 (Dense)             (None, 1024)              1049600
_________________________________________________________________
batch_normalization_10 (Batc (None, 1024)              4096
_________________________________________________________________
activation_9 (Activation)    (None, 1024)              0
_________________________________________________________________
dropout_4 (Dropout)          (None, 1024)              0
_________________________________________________________________
dense_11 (Dense)             (None, 512)               524800
_________________________________________________________________
batch_normalization_11 (Batc (None, 512)               2048
_________________________________________________________________
activation_10 (Activation)   (None, 512)               0
_________________________________________________________________
dense_12 (Dense)             (None, 256)               131328
_________________________________________________________________
batch_normalization_12 (Batc (None, 256)               1024
_________________________________________________________________
activation_11 (Activation)   (None, 256)               0
_________________________________________________________________
dense_13 (Dense)             (None, 128)               32896
_________________________________________________________________
batch_normalization_13 (Batc (None, 128)               512
_________________________________________________________________
activation_12 (Activation)   (None, 128)               0
_________________________________________________________________
dense_14 (Dense)             (None, 64)                8256
_________________________________________________________________
batch_normalization_14 (Batc (None, 64)                256
_________________________________________________________________
activation_13 (Activation)   (None, 64)                0
_________________________________________________________________
dense_15 (Dense)             (None, 32)                2080
_________________________________________________________________
batch_normalization_15 (Batc (None, 32)                128       
_________________________________________________________________
activation_14 (Activation)   (None, 32)                0
_________________________________________________________________
dense_16 (Dense)             (None, 16)                528
_________________________________________________________________
batch_normalization_16 (Batc (None, 16)                64
_________________________________________________________________
activation_15 (Activation)   (None, 16)                0
_________________________________________________________________
dense_17 (Dense)             (None, 8)                 136
_________________________________________________________________
batch_normalization_17 (Batc (None, 8)                 32
_________________________________________________________________
activation_16 (Activation)   (None, 8)                 0
_________________________________________________________________
dense_18 (Dense)             (None, 4)                 36
_________________________________________________________________
batch_normalization_18 (Batc (None, 4)                 16
_________________________________________________________________
activation_17 (Activation)   (None, 4)                 0
_________________________________________________________________
dense_19 (Dense)             (None, 2)                 10
_________________________________________________________________
batch_normalization_19 (Batc (None, 2)                 8
_________________________________________________________________
activation_18 (Activation)   (None, 2)                 0
_________________________________________________________________
dense_20 (Dense)             (None, 1)                 3
=================================================================
Total params: 2,467,113
Trainable params: 2,458,935
Non-trainable params: 8,178
_________________________________________________________________
'''

# 选择优化器
Adagrad = keras.optimizers.Adagrad(lr=0.1, epsilon=None, decay=0.0)

# 设置优化器和损失函数
model.compile(optimizer=Adagrad, loss='mse')

# 导入模型
# model.load_weights(path + 'step=98400&loss=1.9879720564241763.h5')

'''训练'''
less = 100

for step in range(200001):


    loss_train = model.train_on_batch(X_train, Y_train)
    loss_test = model.evaluate(X_test, Y_test, verbose=0)
    
    # print(model.evaluate(X_test,Y_test))
    if step % 100 == 0:

        print("The step is ", step , "..............." + '[loss_train]:', loss_train)
        print("The step is ", step , "..............................................." + '[loss_test]:', loss_test)

    if (step > 10000) & (loss_test < less) :
            
        less = loss_test
        save_path = path + 'step=' + str(step) + '&loss=' + str(loss_test) + '.h5'
        model.save_weights(save_path)


'''保存模型(本次最终)'''
save_path = path + 'loss=' + str(loss_test) + '.h5'
model.save_weights(save_path)



# =================================================================================================================================
# =================================================================================================================================

# 结果记录
'''

'''

替换后

'''在Use3的基础上进行修改, 并使用model.fit()进行训练'''

import csv
from functools import lru_cache
import numpy as np
import matplotlib.pyplot as plt

import keras
from keras import backend as K
from keras.layers import Layer
from keras.layers import Dense
from keras.models import Sequential
from keras.layers import Activation

from keras.callbacks import ReduceLROnPlateau, EarlyStopping, ModelCheckpoint

from pylab import *
from mpl_toolkits.mplot3d import Axes3D


# 重新声明
from keras.layers.advanced_activations import LeakyReLU
class LeakyReLU(LeakyReLU):
    def __init__(self, **kwargs):
        self.__name__ = "LeakyReLU"
        super(LeakyReLU, self).__init__(**kwargs)

leakyrelu_alpha = 0.3

'''设置文件保存路径'''
path = './model_save/4D/Use5/'


'''导入csv文件(返回的数据为ndarry)'''
# 训练集
csv_train = np.loadtxt(open("./train_last.csv","rb"),delimiter=",",skiprows=0) 
print('train.csv导入成功!')
# 测试集
csv_test = np.loadtxt(open("./test_last.csv","rb"),delimiter=",",skiprows=0) 
print('test.csv导入成功!')


'''选取输入向量'''
# 训练集
X_train = csv_train[:,0:3]# [:,np.newaxis]
Y_train = csv_train[:,3]
# 测试集
X_test = csv_test[:,0:3]
Y_test = csv_test[:,3]


# 创建RBF层
class RBFLayer(Layer):
    def __init__(self, units, gamma, **kwargs):
        super(RBFLayer, self).__init__(**kwargs)
        self.units = units
        self.gamma = K.cast_to_floatx(gamma)

    def build(self, input_shape):
        self.mu = self.add_weight(name='mu',
                                  shape=(int(input_shape[1]), self.units),
                                  initializer='uniform',
                                  trainable=True)
        super(RBFLayer, self).build(input_shape)

    def call(self, inputs):
        diff = K.expand_dims(inputs) - self.mu
        l2 = K.sum(K.pow(diff,2), axis=1)
        res = K.exp(-1 * self.gamma * l2)
        return res

    def compute_output_shape(self, input_shape):
        return (input_shape[0], self.units)



'''搭建网络结构'''
'''Glorot 正态分布初始化器，也称为 Xavier 正态分布初始化器。
它从以 0 为中心，标准差为 stddev = sqrt(2 / (fan_in + fan_out)) 的截断正态分布中抽取样本， 
其中 fan_in 是权值张量中的输入单位的数量， fan_out 是权值张量中的输出单位的数量。'''
## keras.initializers.glorot_normal(seed=None) 

model = Sequential()

# 第1层
model.add(Dense(units = 3, input_dim = 3))
model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(Dense(units = 8, input_dim = 3))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 16, input_dim = 8))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 32, input_dim = 16))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 64, input_dim = 32))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 128, input_dim = 64))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 256, input_dim = 128))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 512, input_dim = 256))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 1024, input_dim = 512))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

# model.add(keras.layers.Dropout(0.9))
# model.add(Dense(units = 1024, input_dim = 1024))
# model.add(keras.layers.normalization.BatchNormalization())
# model.add(Activation(keras.layers.advanced_activations.LeakyReLU(alpha=0.3)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 1024, input_dim = 1024))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(keras.layers.Dropout(0.9))
model.add(Dense(units = 512, input_dim = 1024)) 
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 256, input_dim = 512))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 128, input_dim = 256))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 64, input_dim = 128))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 32, input_dim = 64))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 16, input_dim = 32))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 8, input_dim = 16))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 4, input_dim = 8))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 2, input_dim = 4))
model.add(keras.layers.normalization.BatchNormalization())
model.add(Activation(LeakyReLU(alpha=leakyrelu_alpha)))

model.add(Dense(units = 1, input_dim = 64, use_bias=True))
     
# 显示
model.summary()
'''
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_1 (Dense)              (None, 3)                 12
_________________________________________________________________
batch_normalization_1 (Batch (None, 3)                 12
_________________________________________________________________
dense_2 (Dense)              (None, 8)                 32
_________________________________________________________________
batch_normalization_2 (Batch (None, 8)                 32
_________________________________________________________________
activation_1 (Activation)    (None, 8)                 0
_________________________________________________________________
dense_3 (Dense)              (None, 16)                144
_________________________________________________________________
batch_normalization_3 (Batch (None, 16)                64
_________________________________________________________________
activation_2 (Activation)    (None, 16)                0
_________________________________________________________________
dense_4 (Dense)              (None, 32)                544
_________________________________________________________________
batch_normalization_4 (Batch (None, 32)                128
_________________________________________________________________
activation_3 (Activation)    (None, 32)                0
_________________________________________________________________
dense_5 (Dense)              (None, 64)                2112
_________________________________________________________________
batch_normalization_5 (Batch (None, 64)                256
_________________________________________________________________
activation_4 (Activation)    (None, 64)                0
_________________________________________________________________
dense_6 (Dense)              (None, 128)               8320
_________________________________________________________________
batch_normalization_6 (Batch (None, 128)               512
_________________________________________________________________
activation_5 (Activation)    (None, 128)               0
_________________________________________________________________
dense_7 (Dense)              (None, 256)               33024
_________________________________________________________________
batch_normalization_7 (Batch (None, 256)               1024      
_________________________________________________________________
activation_6 (Activation)    (None, 256)               0
_________________________________________________________________
dropout_1 (Dropout)          (None, 256)               0
_________________________________________________________________
dense_8 (Dense)              (None, 512)               131584
_________________________________________________________________
batch_normalization_8 (Batch (None, 512)               2048
_________________________________________________________________
activation_7 (Activation)    (None, 512)               0
_________________________________________________________________
dropout_2 (Dropout)          (None, 512)               0
_________________________________________________________________
dense_9 (Dense)              (None, 1024)              525312
_________________________________________________________________
batch_normalization_9 (Batch (None, 1024)              4096
_________________________________________________________________
activation_8 (Activation)    (None, 1024)              0
_________________________________________________________________
dropout_3 (Dropout)          (None, 1024)              0
_________________________________________________________________
dense_10 (Dense)             (None, 1024)              1049600
_________________________________________________________________
batch_normalization_10 (Batc (None, 1024)              4096
_________________________________________________________________
activation_9 (Activation)    (None, 1024)              0
_________________________________________________________________
dropout_4 (Dropout)          (None, 1024)              0
_________________________________________________________________
dense_11 (Dense)             (None, 512)               524800
_________________________________________________________________
batch_normalization_11 (Batc (None, 512)               2048
_________________________________________________________________
activation_10 (Activation)   (None, 512)               0
_________________________________________________________________
dense_12 (Dense)             (None, 256)               131328
_________________________________________________________________
batch_normalization_12 (Batc (None, 256)               1024
_________________________________________________________________
activation_11 (Activation)   (None, 256)               0
_________________________________________________________________
dense_13 (Dense)             (None, 128)               32896
_________________________________________________________________
batch_normalization_13 (Batc (None, 128)               512
_________________________________________________________________
activation_12 (Activation)   (None, 128)               0
_________________________________________________________________
dense_14 (Dense)             (None, 64)                8256
_________________________________________________________________
batch_normalization_14 (Batc (None, 64)                256
_________________________________________________________________
activation_13 (Activation)   (None, 64)                0
_________________________________________________________________
dense_15 (Dense)             (None, 32)                2080
_________________________________________________________________
batch_normalization_15 (Batc (None, 32)                128       
_________________________________________________________________
activation_14 (Activation)   (None, 32)                0
_________________________________________________________________
dense_16 (Dense)             (None, 16)                528
_________________________________________________________________
batch_normalization_16 (Batc (None, 16)                64
_________________________________________________________________
activation_15 (Activation)   (None, 16)                0
_________________________________________________________________
dense_17 (Dense)             (None, 8)                 136
_________________________________________________________________
batch_normalization_17 (Batc (None, 8)                 32
_________________________________________________________________
activation_16 (Activation)   (None, 8)                 0
_________________________________________________________________
dense_18 (Dense)             (None, 4)                 36
_________________________________________________________________
batch_normalization_18 (Batc (None, 4)                 16
_________________________________________________________________
activation_17 (Activation)   (None, 4)                 0
_________________________________________________________________
dense_19 (Dense)             (None, 2)                 10
_________________________________________________________________
batch_normalization_19 (Batc (None, 2)                 8
_________________________________________________________________
activation_18 (Activation)   (None, 2)                 0
_________________________________________________________________
dense_20 (Dense)             (None, 1)                 3
=================================================================
Total params: 2,467,113
Trainable params: 2,458,935
Non-trainable params: 8,178
_________________________________________________________________
'''

# 选择优化器
Adagrad = keras.optimizers.Adagrad(lr=0.1, epsilon=None, decay=0.0)

# 设置优化器和损失函数
model.compile(optimizer=Adagrad, loss='mse')

# 导入模型
model.load_weights(path + 'weights.hdf5')

'''训练'''
# less = 100

# for step in range(200001):


#     loss_train = model.train_on_batch(X_train, Y_train)
#     loss_test = model.evaluate(X_test, Y_test, verbose=0)
    
#     # print(model.evaluate(X_test,Y_test))
#     if step % 100 == 0:

#         print("The step is ", step , "..............." + '[loss_train]:', loss_train)
#         print("The step is ", step , "..............................................." + '[loss_test]:', loss_test)

#     if (step > 10000) & (loss_test < less) :
            
#         less = loss_test
#         save_path = path + 'step=' + str(step) + '&loss=' + str(loss_test) + '.h5'
#         model.save_weights(save_path)

# 模型不收敛则停止
EarlyStop=EarlyStopping(monitor='mse',
                        patience=2,
                        verbose=1, 
                        mode='auto')
#减小学习率
Reduce=ReduceLROnPlateau(monitor='mse',
                         factor=0.1,
                         patience=1,
                         verbose=1,
                         mode='auto',
                         epsilon=0.0001,
                         cooldown=0,
                         min_lr=0)
# 保存模型
checkpointer = ModelCheckpoint(filepath = path + 'weights.hdf5', 
                               verbose=1, 
                               save_best_only=True)

# Fit the model
history=model.fit(X_train, Y_train, 
                  validation_data=(X_test, Y_test),
                  epochs=200, batch_size=200,
                  callbacks=[EarlyStop, Reduce, checkpointer], 
                  verbose=2)

# Final evaluation of the model
scores = model.evaluate(X_test, Y_test, verbose=1)
# print("Baseline Error: %.2f%%" % (100-scores[1]*100))
print(scores)


'''保存模型(本次最终)'''
save_path = path + 'loss=' + str(scores) + '.h5'
model.save_weights(save_path)



# =================================================================================================================================
# =================================================================================================================================

# 结果记录
'''

'''
Leisureconfused
关注
  • 0
    点赞
  • 踩
  • 2
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
在keras中model.fit_generator()和model.fit()的区别说明
09-16
主要介绍了在keras中model.fit_generator()和model.fit()的区别说明，具有很好的参考价值，希望对大家有所帮助。一起跟随小编过来看看吧
keras model.fit 解决validation_spilt=num 的问题
09-16
主要介绍了keras model.fit 解决validation_spilt=num 的问题，具有很好的参考价值，希望对大家有所帮助。一起跟随小编过来看看吧
解决Tensorflow2.0 tf.keras.Model.load_weights() 报错处理问题
09-16
主要介绍了解决Tensorflow2.0 tf.keras.Model.load_weights() 报错处理问题，具有很好的参考价值，希望对大家有所帮助。一起跟随想过来看看吧
keras 自定义loss model.add_loss的使用详解
09-16
主要介绍了keras 自定义loss model.add_loss的使用详解，具有很好的参考价值，希望对大家有所帮助。一起跟随小编过来看看吧
gcForest+++xgboost的树木的个数++train和fit的区别
lusic01的专栏
01-10 1万+
xgboost 逻辑回归：objective参数（reg:logistic,binary:logistic,binary:logitraw,）对比分析 置顶 2018年07月11日 22:26:53 phyllisyuell 阅读数：1755 一、问题 熟悉xgboost的小伙伴都知道，它在训练模型时，有train（）方法和fit（）方法，两种方法都是用于构建模型的，然而在使用过程中有什么不...
model.train_on_batch介绍
weixin_47272689的博客
09-16 964
train_on_batch
Keras学习（二）—— 模型相关函数: model.train_on_batch()
Work hard
08-21 2万+
model.train_on_batch 函数作用函数定义 函数作用 model.train_on_batch() 在训练集数据的一批数据上进行训练 函数定义 train_on_batch(x, y, sample_weight=None, class_weight=None) 参数含义： x Numpy训练数据数组，或Numpy数组列表（如果模型有多个输入）。 如果模型中的所有输入都已...
keras train_on_batch详解（train_on_batch的输出输入详解，train_on_batch多GPU训练详解，自定义学习率调整策略）
热门推荐
baoxin1100的博客
08-11 2万+
利用 train_on_batch 精细管理训练过程 大部分使用 keras 的同学使用 fit() 或者 fit_generator() 进行模型训练， 这两个 api 对于刚接触深度学习的同学非常友好和方便，但是由于其是非常深度的封装，对于希望自定义训练过程的同学就显得不是那么方便（从 torch 转 keras 的同学可能更喜欢自定义训练过程），而且，对于 GAN 这种需要分步进行训练的模型，也无法直接使用 fit 或者 fit_generator 直接训练的。因此，keras 提供了 train_o
Keras之model.compile与train_on_batch和TensorBoard for train_on_batch
鹊踏枝-码农的专栏
11-10 1万+
在利用TensorFlow的TensorBoard对train_on_batch的输出进行画图时发现了一些问题。下面对train_on_batch的输出进行讲解。在讲解train_on_batch之前，先看一下Keras的model.compile函数。下面利用Keras版Faster R-CNN代码进行讲解。示例代码如下： # define the RPN, built on the bas...
tensorflow中的batch数据打印
shenjianhua005的专栏
07-09 865
之前一直困惑怎么打印batch返回值，后来发现一个迭代就可以了，因为这个batch返回的就是迭代器。 首先 batch 就是把数据分成几组，前面己组的数据大小都是batch的大小，但是最后一组可以能数据量不多，就只有一点 如下，先创建一个20行3列的数据 temp = tf.Variable(tf.random.normal(shape=(20, 3), mean=0, stddev=0.01, dtype=tf.float32)) print("temp") print(temp) 然后按照bat
kera中train_on_batch 自定义训练过程
爱CV
03-22 3168
train_on_batch 可以在keras中自定义精细化训练过程使用。 使用示例： import os os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' import numpy as np import matplotlib.pyplot as plt from keras.models import Sequential from keras.lay...
keras在train_on_batch中使用generator
fun_always的博客
07-31 855
Generator是keras中很方便的数据输入方式，既可以节省内存空间，又自带数据增强的功能，一般用于fit_generator这种比较单一的训练方式，不适于train_on_batch这种拓展性较高的训练方式。但实际上generator是可以用于train_on_batch的，下面介绍具体方法： 理解generator 理解generator实际上理解yield关键词就够了，yield可以看作一个带指针的return，每次返回时指针指向程序停止的位置，因此下一次可以接着上一次运行。在外面调用gene
[TensorFlow 2] [Keras] fit()、fit_generator() 和 train_on_batch() 分析与应用
Xovee
06-09 1万+
本文主要内容：Keras 中的`fit()`函数、`fit_generator()`函数、`train_on_batch()`函数的分析及应用。
【Python-Keras】Keras搭建神经网络模型的Model解析与使用
BetterBench的博客
01-05 4799
1 作用 用于搭建和配置神经网络训练的模型，通过Model这个方法可以调用很多API去实现训练神经网络。 2 API介绍 （1）Model.fit() 参数： Model.fit( x=None, y=None, batch_size=None, epochs=1, verbose=1, callbacks=None, validation_split=0.0, validation_data=None, shuffle=True,
TensorFlow2.0学习笔记-3.模型训练
ruoqi23的博客
08-06 2254
3.模型训练 3.1.Keras版本模型训练 •构建模型（顺序模型、函数式模型、子类模型） •模型训练：model.fit() •模型验证：model.evaluate() •模型预测：model.predict() •使用样本加权和类别加权 回调函数 • ModelCheckpoint：定期保存模型。 • EarlyStopping：当训练不再改善验证指标时，停止培训。 • TensorBoard：定期编写可在TensorBoard中可视化的模型日志（更多详细信息，请参见“可视化”...
1.神经网络基础
乌龟慢慢爬
03-13 811
1.感知机就是一个很简单的模型，给定输入，输出0或12.多层感知机（MLP）网络是密集的（dense），意味着一个层中的每一个神经元都会与前一层的每个神经元相连接，并且连接到下一层中的所有神经元。3.训练感知机的问题以及应对之道我们希望我们的学习是一点一点进步的，而不是跳跃的，对于下面函数来说，它的输出有一个很大的跳跃，不能够逐步学习。所以我们需要一个函数，从0逐步变化到1，没有间断。从数学上讲，...
Keras框架常见问题
pengxiang1998的博客
09-25 944
Keras框架使用常见问题。
使用model.fit_generator方法进行训练(自己的训练集-多分类)
xfjjs_net的博客
12-04 2万+
我们在使用model.fit()进行训练的时候， 在这之前你肯定会有训练集的x_img_train,y_label_train两个参数。 fit(x=None, y=None, batch_size=None, epochs=1, verbose=1, callbacks=None, validation_split=0.0, validation_data=None, shuffle=Tru...
model.train_on_batch
最新发布
11-03
model.train_on_batch是Keras中的一个训练函数，它用于在一个batch的训练数据上对模型进行训练。在这个函数中，首先把训练数据按照batch_size分批次加载，然后利用这些数据进行模型的前向传播和反向传播，并根据优化算法更新模型的参数，使得模型能够更好地拟合训练数据。 具体而言，model.train_on_batch的具体步骤如下： 1. 通过调用model的compile方法对模型进行配置，包括选择损失函数、优化器等等，以便接下来的训练能够按照指定的方式进行。 2. 加载一个batch的训练数据，包括输入数据和对应的标签。 3. 调用model的train_on_batch方法来进行训练，该方法会执行前向传播、计算损失、反向传播、参数更新等操作。 4. 返回当前batch的训练损失值，可以用于对训练过程进行监控和评估。 5. 重复执行步骤2至4，直到所有的训练数据都被用于训练。 需要注意的是，model.train_on_batch是一次性训练一个batch的数据，并且不会返回整个训练过程的性能指标，如准确率和损失值的变化。如果需要对整个训练过程进行监控和评估，可以使用其他的训练函数，如model.fit。 总的来说，model.train_on_batch是Keras中用于对模型进行训练的函数，可以有效地利用大量的训练数据进行模型的更新和参数优化，从而提高模型的拟合能力和性能。

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值