Skip to content
Related Articles

Related Articles

Tensorflow.js tf.layers.convLstm2d() Function

Improve Article
Save Article
  • Last Updated : 30 May, 2022
Improve Article
Save Article

Tensorflow.js is an open-source library that is being developed by Google for running machine learning models as well as deep learning neural networks in the browser or node environment.

The tf.layers.convLstm2d() function is used for creating a ConvRNN2D layer which consists of one ConvLSTM2DCell and the apply method of ConvLSTM2D operates on a sequence of inputs. The shape of the input needs to be at least 4D with the first dimension being time steps. convLstm2d is the Convolutional LSTM layer.

Syntax:

tf.layers.convLstm2d( args )

Parameters:

  • args: It specifies the given config object:
    • activation: It is the activation function that is to be used. Default is a hyperbolic tangent. If null, is passed then no activation function will be used.
    • useBias: It specifies whether to use a bias vector or not.
    • kernelInitializer: It is the Initializer of the kernel weight matrix, which is used for linear transformation of the inputs.
    • recurrentInitializer: It is the Initializer of the recurrentInitializer weight matrix, which is used for linear transformation of the recurrent state.
    • biasInitializer: It is the initializer for the bias vector.
    • kernelRegularizer: It is a regularizer function applied to the kernel weight matrix. 
    • recurrentRegularizer: It is a regularizer function applied to the recurrentKernel weight matrix.
    • biasRegularizer:  It is a regularizer function applied to the bias vector.
    • kernelConstraint: It is a constraint function applied to the kernel weight matrix.
    • recurrentConstraint:  It is a constraint function applied to the recurrentKernel weight matrix.
    • biasConstraint: It is a constraint function applied to the bias vector.
    • dropout: It is a number between 0 to 1 which indicate the unit to drop for the linear transformation of the input.
    • recurrentDropout: It is the fraction number that drops the linear transformation of the recurrent state.
    • dropoutFunc: This is used for the test DI. 
    • inputShape: It should be an array of numbers. This field is used to create an input layer which is used to be inserted before this layer.
    • batchInputShape: It should be an array of numbers. This field will be used if inputShape and this field are provided as a parameter for creating the input layer which is used to insert before this layer.
    • batchSize: It should be a number. In the absence of batchInputShape this field is used to create batchInputShape with inputShape. batchInputShape : [ batchSize ,  …inputShape].
    • dtype: If this layer is used as the input layer, then this field is used as data type for this layer.
    • name: It should be a string type. this field defines the name for this layer.
    • trainable: It should be boolean. This field defines whether the weights of this layer are trainable with fit or not.
    • weights: This should be a tensor that defines the initial weight value for this layer.
    • inputDType: This is a data type that is used for Legacy support.
    • recurrentActivation: It specifies the activation function which will be used for the recurrent step. The default value of this parameter is hard sigmoid.
    • unitForgetBias: It should be boolean. It is used to add 1 to the bias of the forget gate at initialization. 
    • implementation: It specifies the implementation mode. It can be either 1 or 2. For superior performance, implementation is recommended.
    • returnSequences:  It should be boolean. It defines which both of these weather last output or full sequence return in the output sequence.
    • returnState: It should be boolean. It defines the return last state in output.
    • goBackwards: It should be boolean. It defines whether to process the input in reverse and reverse sequence or not.
    • stateful: It should be boolean. It defines whether to use the last state of the batch as the initial state for the current batch or not.
    • unroll: It should be boolean. It tells whether to use unrolled network else, use a symbolic loop. 
    • inputDim: It should be a number. This is used when this layer is used as the first layer of the model. It defines the dimension of input in the layer. 
    • inputLength: It should be a number. This field should be when the sequence in input data is constant. 
    • cell: It should be an instance of RNN cell or an array of instances of RNN cells.
    • filters: It is a number of filters in convolution.
    • kernelSize: It is a dimension that the convolution window will have.
    • strides: It is strides of convolution layer in each dimension. 
    • padding: It should be ‘valid’, ‘same’, and ‘causal’. It defines the padding mode. 
    • dataFormat: It defines the format of data, which tells the ordering of the dimensions in the inputs.
    • dilationRate: The dilation rate to which the convolution layer should dilate in each dimension.

Returns: It returns the ConvLSTM2D.

Example 1: In this example, we will operate our apply method in only one input.

Javascript




// Config for input layer
const bSize = 3;
const sLength = 4;
const size = 7;
const channels = 2;
 
const inputShape = [bSize, sLength, size, size, channels];
const input = tf.zeros(inputShape);
 
// Config for convLstm2d
const filters = 4;
const kernelSize = 4;
 
const layer = tf.layers.convLstm2d({filters, kernelSize});
 
const output = layer.apply(input);
 
// Printing Shape of our output
console.log(JSON.stringify(output.shape));


Output:

[3,4,4,4]

Example 2:

Javascript




import * as tf form "@tensorflow/tfjs"
 
const filters = 3;
const kernelSize = 3;
 
const batchSize = 4;
const sequenceLength = 2;
const size = 5;
const channels = 3;
 
const Input = tf.input({shape: [2, 3, 5, 4]});
const Data = tf.ones([1, 2, 3, 5, 4]);
const inputShape = [batchSize,
    sequenceLength, size, size, channels];
const input = tf.ones(inputShape);
 
const layer = tf.layers.convLstm2d({filters, kernelSize});
 
const output = layer.apply(Input);
const model = tf.model({inputs: Input, outputs: output});
 
model.predict(Data).print();


Output:

‚ÄčTensor
    [[[[0.2382042, 0.3073192, -0.0167814],
       [0.2362143, 0.3104056, -0.0188109],
       [0.2277206, 0.3124302, -0.0153713]]]]

Reference: https://js.tensorflow.org/api/latest/#layers.convLstm2d


My Personal Notes arrow_drop_up
Related Articles

Start Your Coding Journey Now!