Tensorflow.js tf.layers.dense() Function

The tf.layers.dense() is an inbuilt function of Tensorflow.js library. This function is used to create fully connected layers, in which every output depends on every input.

Syntax:

tf.layers.dense(args)

Parameters: This function takes the args object as a parameter which can have the following properties:

• units: It is a positive number that defines the dimensionality of the output space.
• activation: Specifies which activation function to use.
• useBias: specifies whether to apply a bias or not.
• kernelInitializer: Specifies which initializer to use for the dense kernel weight matrix.
• biasInitializer: Specifies the bias vector for this layer.
• inputDim: Defines input shape as [inputDim].
• kernelConstraint: Specifies the constraint for the kernel.
• biasConstraint: Specifics constraint for the bias vector.
• kernelRegularizer: Specifies the regularizer function applied to the dense kernel weights matrix.
• biasRegularizer: Specifies the regularizer function applied to the bias vector.
• activityRegularizer: Specifies the regularizer function applied to the activation.
• inputShape: If this parameter is defined, it will create another input layer to insert before this layer.
• batchInputShape: If this parameter is defined, it will create another input layer to insert before this layer.
• batchSize : Used to construct batchInputShape, if not already specified.
• dtype: Specifies the data type for this layer. The default value of this parameter is ‘float32’.
• name: Specifies name for this layer.
• trainable: Specifies whether the weights of this layer are updated by fit.
• weights: Specifies the initial weight values of the layer.
• inputDType : It is used to denote the inputDType and its value can be ‘float32’ or ‘int32’ or ‘bool’ or ‘complex64’ or ‘string’.

Return value: It returns a Dense object.

Example 1:

Javascript

import * as tf from "@tensorflow/tfjs"

// Create a new dense layer
const denseLayer = tf.layers.dense({
   units: 2,
   kernelInitializer: 'heNormal',
   useBias: true
});

const input = tf.ones([2, 3]);
const output = denseLayer.apply(input);

// Print the output
output.print()

Output:

Example 2:

Javascript

import * as tf from "@tensorflow/tfjs"

// Create a new dense layer
const denseLayer = tf.layers.dense({
   units: 3,
   kernelInitializer: 'heNormal',
   useBias: false
});

const input = tf.ones([2, 3, 3]);
const output = denseLayer.apply(input);

// Print the output
output.print()

Output:

Example 3:

Javascript

import * as tf from "@tensorflow/tfjs"

// Create a new dense layer
const denseLayer = tf.layers.dense({
   units: 3,
   kernelInitializer: 'ones',
   useBias: false
});

const input = tf.ones([2, 3, 3]);
const output = denseLayer.apply(input);

// Print the output
output.print()

Output:

Example 4:

Javascript

import * as tf from "@tensorflow/tfjs"

// Create a new dense layer
const denseLayer = tf.layers.dense({
   units: 3,
   kernelInitializer: 'randomUniform',
   useBias: false
});

const input = tf.ones([2, 3, 3]);
const output = denseLayer.apply(input);

// Print the output
output.print()

Output:

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