确切模型在keras-tf上收敛,但在keras上不收敛

时间:2019-08-07 14:15:22

标签: python tensorflow keras deep-learning

我正在使用简单的RNN预测时间序列上的EWMA (exponential weighted moving average) formula。已经发布了关于它的here

虽然使用keras-tf(来自tensorflow导入keras)完美地收敛了模型,但使用本地keras(导入keras)无法完全相同的代码。

融合模型代码(keras-tf):

from tensorflow import keras
import numpy as np

np.random.seed(1337)  # for reproducibility

def run_avg(signal, alpha=0.2):
    avg_signal = []
    avg = np.mean(signal)
    for i, sample in enumerate(signal):
        if np.isnan(sample) or sample == 0:
            sample = avg
        avg = (1 - alpha) * avg + alpha * sample
        avg_signal.append(avg)
    return np.array(avg_signal)

def train():
    x = np.random.rand(3000)
    y = run_avg(x)
    x = np.reshape(x, (-1, 1, 1))
    y = np.reshape(y, (-1, 1))

    input_layer = keras.layers.Input(batch_shape=(1, 1, 1), dtype='float32')
    rnn_layer = keras.layers.SimpleRNN(1, stateful=True, activation=None, name='rnn_layer_1')(input_layer)
    model = keras.Model(inputs=input_layer, outputs=rnn_layer)

    model.compile(optimizer=keras.optimizers.SGD(lr=0.1), loss='mse')
    model.summary()

    print(model.get_layer('rnn_layer_1').get_weights())
    model.fit(x=x, y=y, batch_size=1, epochs=10, shuffle=False)
    print(model.get_layer('rnn_layer_1').get_weights())

train()

没有收敛的模型代码:

from keras import Model
from keras.layers import SimpleRNN, Input
from keras.optimizers import SGD

import numpy as np

np.random.seed(1337)  # for reproducibility

def run_avg(signal, alpha=0.2):
    avg_signal = []
    avg = np.mean(signal)
    for i, sample in enumerate(signal):
        if np.isnan(sample) or sample == 0:
            sample = avg
        avg = (1 - alpha) * avg + alpha * sample
        avg_signal.append(avg)
    return np.array(avg_signal)

def train():
    x = np.random.rand(3000)
    y = run_avg(x)
    x = np.reshape(x, (-1, 1, 1))
    y = np.reshape(y, (-1, 1))

    input_layer = Input(batch_shape=(1, 1, 1), dtype='float32')
    rnn_layer = SimpleRNN(1, stateful=True, activation=None, name='rnn_layer_1')(input_layer)
    model = Model(inputs=input_layer, outputs=rnn_layer)


    model.compile(optimizer=SGD(lr=0.1), loss='mse')
    model.summary()

    print(model.get_layer('rnn_layer_1').get_weights())
    model.fit(x=x, y=y, batch_size=1, epochs=10, shuffle=False)
    print(model.get_layer('rnn_layer_1').get_weights())

train()

在tf-keras收敛模型中,损耗最小化,权重近似为EWMA公式, 在非收敛模型中,损失激增至nan。 据我所知,唯一的区别是导入类的方式。

我在两种实现方式中都使用了相同的随机种子。我正在Windows PC,带有keras 2.2.4和tensorflow版本1.13.1(其中包括版本2.2.4-tf中的keras)的anaconda环境中工作。

对此有何见解?

1 个答案:

答案 0 :(得分:1)

这可能是由于TF KerasNative Keras之间在SimpleRNN的实现上存在差异(1个衬里)。

下面提到的行在TF Keras中实现,而在Keras中不实现。

self.input_spec = [InputSpec(ndim=3)]

这种差异的一种情况就是您上面提到的情况。

我想使用Sequential类的Keras来演示类似的情况。

以下代码对TF Keras来说很有效:

from tensorflow import keras
import numpy as np
from tensorflow.keras.models import Sequential as Sequential

np.random.seed(1337)  # for reproducibility

def run_avg(signal, alpha=0.2):
    avg_signal = []
    avg = np.mean(signal)
    for i, sample in enumerate(signal):
        if np.isnan(sample) or sample == 0:
            sample = avg
        avg = (1 - alpha) * avg + alpha * sample
        avg_signal.append(avg)
    return np.array(avg_signal)

def train():
    x = np.random.rand(3000)
    y = run_avg(x)
    x = np.reshape(x, (-1, 1, 1))
    y = np.reshape(y, (-1, 1))

    # SimpleRNN model
    model = Sequential()
    model.add(keras.layers.Input(batch_shape=(1, 1, 1), dtype='float32'))
    model.add(keras.layers.SimpleRNN(1, stateful=True, activation=None, name='rnn_layer_1'))
    model.compile(optimizer=keras.optimizers.SGD(lr=0.1), loss='mse')
    model.summary()

    print(model.get_layer('rnn_layer_1').get_weights())
    model.fit(x=x, y=y, batch_size=1, epochs=10, shuffle=False)
    print(model.get_layer('rnn_layer_1').get_weights())

train()

但是,如果我们使用Native Keras运行相同的代码,则会显示以下错误:

TypeError: The added layer must be an instance of class Layer. Found: Tensor("input_1_1:0", shape=(1, 1, 1), dtype=float32)

如果我们替换下面的代码行

model.add(Input(batch_shape=(1, 1, 1), dtype='float32'))

带有以下代码,

model.add(Dense(32, batch_input_shape=(1,1,1), dtype='float32'))

具有model实现的Keras甚至与TF Keras实现的收敛相似。

在两种情况下,如果您想从代码角度了解实现的差异,可以参考以下链接:

https://github.com/tensorflow/tensorflow/blob/r1.14/tensorflow/python/keras/layers/recurrent.py#L1364-L1375

https://github.com/keras-team/keras/blob/master/keras/layers/recurrent.py#L1082-L1091

如果您认为此答案有用,请接受此答案和/或对其进行投票。谢谢。