Keras LAMBDA Layer

Lambda layer

Keras.layers.core.Lambda (function, Output_shape=none, Mask=none, Arguments=none)

This function is used to apply any theano/tensorflow expression to the output of the previous layer

If you just want to make a change to the data that flows through the layer, and the transformation itself has no parameters to learn, then it is most appropriate to use a lambda layer directly.

The Import method is

 from Import Lambda

The lambda function accepts two parameters, the first one is the mapping function of the input tensor to the output tensor, and the second is the mapping function of the shape to the output shape of the input.

Parameters

• Functions: A function to be implemented that accepts only one variable, that is, the output of the previous layer

• Output_shape: The shape of the value that the function should return, can be a tuple, or a function that calculates the output shape from the input shape

• Mask: Masking film

• Arguments: optional, dictionary, used to record additional keyword arguments passed to a function

Example

#Add a X-x^2 layerModel.add (Lambda (LambdaX:X * * 2))#add a layer that returns the concatenation# of the positive part of the the input and#The opposite of the negative partdefantirectifier (x): x-= K.mean (x, Axis=1, keepdims=True) x= K.l2_normalize (x, Axis=1) Pos=k.relu (x) Neg= K.relu (-x)returnK.concatenate ([Pos, neg], Axis=1)defAntirectifier_output_shape (input_shape): Shape=list (input_shape)assertLen (shape) = = 2#Only valid for 2D tensorsShape[-1] *= 2returntuple (Shape) model.add (Lambda (antirectifier, Output_shape=antirectifier_output_shape))

Enter shape

Arbitrarily, when using this layer as the first layer, specify the input_shape

Output shape

The output shape specified by the output_shape parameter, which can be inferred automatically when using TensorFlow

================================================

Keras Lambda Custom layer implementation of data slicing, lambda parameter 1, the code is as follows:

ImportNumPy as NP fromKeras.modelsImportSequential fromKeras.layersImportDense, Activation,reshape fromKeras.layersImportMerge fromKeras.utils.visualize_utilImportplot fromKeras.layersImportInput, Lambda fromKeras.modelsImportModeldefSlice (x,index):returnX[:,:,index]a= Input (shape= (4,2)) X1= Lambda (slice,output_shape= (4,1), arguments={'Index': 0}) (a) X2= Lambda (slice,output_shape= (4,1), arguments={'Index': 1}) (a) X1= Reshape ((4,1,1) ) (x1) X2= Reshape ((4,1,1) ) (x2) Output= Merge ([x1,x2],mode='concat')
Model=Model (A, output) X_test= Np.array ([[[[1,2],[2,3],[3,4],[4,5]]])Printmodel.predict (x_test) plot (model, To_file='Lambda.png', show_shapes=true)

2. Note that lambda can be passed as a parameter, as described in the following code:

def Slice (x,index):     return X[:,:,index]

as above, index is the parameter, passing in the argument through the dictionary.

X1 = Lambda (slice,output_shape= (4,1), arguments={'index'= lambda (slice,output_ Shape= (4,1), arguments={'index': 1}) (a)

3, the above code is implemented, each column of the matrix is extracted, and then operate separately, and finally in a spell together. The visualization diagram is shown below.

Reference:

53414068

54936185

https://keras-cn.readthedocs.io/en/latest/layers/core_layer/

From for notes (Wiz)

Keras LAMBDA Layer