In Keras, what are the layers(functions) corresponding to tf.nn.conv2d_transpose in Tensorflow? I once saw the comment that we can Just use combinations of UpSampling2D and Convolution2D as appropriate. Is that right?
在Keras中，对应于tfnn的层(函数)是什么。在Tensorflow conv2d_transpose吗?我曾经看到过这样的评论，我们可以适当地使用UpSampling2D和卷积2d的组合。是这样吗?
In the following two examples, they all use this kind of combination.
在下面两个例子中，他们都使用这种组合。
1) In Building Autoencoders in Keras, author builds decoder as follows.
1)在Keras中构建Autoencoders时，作者构建了如下的解码器。
2) In an u-uet implementation, author builds deconvolution as follows
2)在u-uet实现中，作者建立了如下的反褶积。
up6 = merge([UpSampling2D(size=(2, 2))(conv5), conv4], mode='concat', concat_axis=1)
conv6 = Convolution2D(256, 3, 3, activation='relu', border_mode='same')(up6)
conv6 = Convolution2D(256, 3, 3, activation='relu', border_mode='same')(conv6)
1 个解决方案
#1
2
The corresponding layers in Keras are Deconvolution2D layers.
在Keras中相应的层是反卷积2d层。
It's worth to mention that you should be really careful with them because they sometimes might behave in unexpected way. I strongly advise you to read this Stack Overflow question (and its answer) before you start to use this layer.
值得一提的是，你应该对他们非常小心，因为他们有时会表现得出乎意料。在开始使用这一层之前，我强烈建议您阅读这个堆栈溢出问题(及其答案)。
UPDATE:
更新:
- Deconvolution is a layer which was add relatively recently - and maybe this is the reason why people advise you to use
Convolution2D * UpSampling2D. - 反褶积是最近添加的一层，也许这就是为什么人们建议你使用卷积2d * UpSampling2D的原因。
- Because it's relatively new - it may not work correctly in some cases. It also need some experience to use them properly.
- 因为它是相对较新的——在某些情况下它可能不能正常工作。它还需要一些经验来正确地使用它们。
- In fact - from a mathematical point of view - every Deconvolution might be presented as a composition of
Convolution2DandUpSampling2D- so maybe this is the reason why it was mentioned in texts you provided. - 事实上，从数学的角度来看，每一次反褶积都可能是由卷积2d和UpSampling2D构成的，所以这可能就是你提供的文本中提到它的原因。
UPDATE 2:
更新2:
Ok. I think I found an easy explaination why Deconvolution2D might be presented in a form of a composition of Convolution2D and UpSampling2D. We would use a definition that Deconvolution2D is a gradient of some convolution layer. Let's consider three most common cases:
好的。我想我找到了一个简单的解释，为什么解卷积2d可以以卷积2d和UpSampling2D的形式呈现。我们会用一个定义来解卷积二维是一个卷积层的梯度。让我们考虑三个最常见的案例:
- The easiest one is a
Convolutional2Dwithout any pooling. In this case - as it's the linear operation - its gradient is a function itself - soConvolution2D. - 最简单的方法是不使用任何池的卷积。在这种情况下——因为它是线性操作——它的梯度是一个函数本身——所以是卷积2d。
- The more tricky one is a gradient of
Convolution2DwithAveragePooling. So:(AveragePooling2D * Convolution2D)' = AveragePooling2D' * Convolution2D'. But a gradient ofAveragePooling2D = UpSample2D * constant- so it's also in this case when the preposition is true. - 比较棘手的一个是卷积2d和平均池的梯度。所以:(AveragePooling2D *卷积2d)' = AveragePooling2D' *卷积2d '。但是，平均pooling2d = UpSample2D *常数的梯度，所以在这种情况下，当介词为真时也是如此。
- The most tricky one is one with
MaxPooling2D. In this case still(MaxPooling2D * Convolution2D)' = MaxPooling2D' * Convolution2D'ButMaxPooling2D' != UpSample2D. But in this case one can easily find an easyConvolution2Dwhich makesMaxPooling2D' = Convolution2D * UpSample2D(intuitively - a gradient ofMaxPooling2Dis a zero matrix with only one 1 on its diagonal. AsConvolution2Dmight express a matrix operation - it may also represent the injection from a identity matrix to aMaxPooling2Dgradient). So:(MaxPooling2D * Convolution2D)' = UpSampling2D * Convolution2D * Convolution2D = UpSampling2D * Convolution2D'. - 最棘手的是使用MaxPooling2D。在这种情况下，仍然(MaxPooling2D *卷积2d)' = MaxPooling2D' *卷积2d '，但MaxPooling2D' != UpSample2D。但在这种情况下，我们可以很容易地找到一个简单的卷积2d，它使MaxPooling2D' =卷积2d * UpSample2D(直观地——MaxPooling2D的梯度是一个零矩阵，在对角线上只有一个1)。由于卷积2d可能表示矩阵运算，它也可以表示从单位矩阵到MaxPooling2D梯度的注入。所以:(MaxPooling2D *卷积2d)' = UpSampling2D *卷积2d = UpSampling2D *卷积2d '。
The final remark is that all parts of the proof have shown that Deconvolution2D is a composition of UpSampling2D and Convolution2D instead of opposite. One can easily proof that every function of a form a composition of UpSampling2D and Convolution2D might be easily presented in a form of a composition of UpSampling2D and Convolution2D. So basically - the proof is done :)
最后的结论是，证明的所有部分都表明，反卷积2d是一种UpSampling2D和卷积2d的合成，而不是相反的。我们可以很容易地证明，一个构成UpSampling2D和卷积2d的每一个函数都可以很容易地以UpSampling2D和卷积2d的形式呈现出来。基本上，证明是这样的
#1
2
The corresponding layers in Keras are Deconvolution2D layers.
在Keras中相应的层是反卷积2d层。
It's worth to mention that you should be really careful with them because they sometimes might behave in unexpected way. I strongly advise you to read this Stack Overflow question (and its answer) before you start to use this layer.
值得一提的是，你应该对他们非常小心，因为他们有时会表现得出乎意料。在开始使用这一层之前，我强烈建议您阅读这个堆栈溢出问题(及其答案)。
UPDATE:
更新:
- Deconvolution is a layer which was add relatively recently - and maybe this is the reason why people advise you to use
Convolution2D * UpSampling2D. - 反褶积是最近添加的一层，也许这就是为什么人们建议你使用卷积2d * UpSampling2D的原因。
- Because it's relatively new - it may not work correctly in some cases. It also need some experience to use them properly.
- 因为它是相对较新的——在某些情况下它可能不能正常工作。它还需要一些经验来正确地使用它们。
- In fact - from a mathematical point of view - every Deconvolution might be presented as a composition of
Convolution2DandUpSampling2D- so maybe this is the reason why it was mentioned in texts you provided. - 事实上，从数学的角度来看，每一次反褶积都可能是由卷积2d和UpSampling2D构成的，所以这可能就是你提供的文本中提到它的原因。
UPDATE 2:
更新2:
Ok. I think I found an easy explaination why Deconvolution2D might be presented in a form of a composition of Convolution2D and UpSampling2D. We would use a definition that Deconvolution2D is a gradient of some convolution layer. Let's consider three most common cases:
好的。我想我找到了一个简单的解释，为什么解卷积2d可以以卷积2d和UpSampling2D的形式呈现。我们会用一个定义来解卷积二维是一个卷积层的梯度。让我们考虑三个最常见的案例:
- The easiest one is a
Convolutional2Dwithout any pooling. In this case - as it's the linear operation - its gradient is a function itself - soConvolution2D. - 最简单的方法是不使用任何池的卷积。在这种情况下——因为它是线性操作——它的梯度是一个函数本身——所以是卷积2d。
- The more tricky one is a gradient of
Convolution2DwithAveragePooling. So:(AveragePooling2D * Convolution2D)' = AveragePooling2D' * Convolution2D'. But a gradient ofAveragePooling2D = UpSample2D * constant- so it's also in this case when the preposition is true. - 比较棘手的一个是卷积2d和平均池的梯度。所以:(AveragePooling2D *卷积2d)' = AveragePooling2D' *卷积2d '。但是，平均pooling2d = UpSample2D *常数的梯度，所以在这种情况下，当介词为真时也是如此。
- The most tricky one is one with
MaxPooling2D. In this case still(MaxPooling2D * Convolution2D)' = MaxPooling2D' * Convolution2D'ButMaxPooling2D' != UpSample2D. But in this case one can easily find an easyConvolution2Dwhich makesMaxPooling2D' = Convolution2D * UpSample2D(intuitively - a gradient ofMaxPooling2Dis a zero matrix with only one 1 on its diagonal. AsConvolution2Dmight express a matrix operation - it may also represent the injection from a identity matrix to aMaxPooling2Dgradient). So:(MaxPooling2D * Convolution2D)' = UpSampling2D * Convolution2D * Convolution2D = UpSampling2D * Convolution2D'. - 最棘手的是使用MaxPooling2D。在这种情况下，仍然(MaxPooling2D *卷积2d)' = MaxPooling2D' *卷积2d '，但MaxPooling2D' != UpSample2D。但在这种情况下，我们可以很容易地找到一个简单的卷积2d，它使MaxPooling2D' =卷积2d * UpSample2D(直观地——MaxPooling2D的梯度是一个零矩阵，在对角线上只有一个1)。由于卷积2d可能表示矩阵运算，它也可以表示从单位矩阵到MaxPooling2D梯度的注入。所以:(MaxPooling2D *卷积2d)' = UpSampling2D *卷积2d = UpSampling2D *卷积2d '。
The final remark is that all parts of the proof have shown that Deconvolution2D is a composition of UpSampling2D and Convolution2D instead of opposite. One can easily proof that every function of a form a composition of UpSampling2D and Convolution2D might be easily presented in a form of a composition of UpSampling2D and Convolution2D. So basically - the proof is done :)
最后的结论是，证明的所有部分都表明，反卷积2d是一种UpSampling2D和卷积2d的合成，而不是相反的。我们可以很容易地证明，一个构成UpSampling2D和卷积2d的每一个函数都可以很容易地以UpSampling2D和卷积2d的形式呈现出来。基本上，证明是这样的