Tensorflow.js tf.sparseToDense() Function

Last Updated : 24 May, 2021

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

The tf.sparseToDense() function is used to convert a specified sparse representation into a dense Tensor. If the given indices get repeated, the final value is summed over all values for that indices.

Syntax:

tf.sparseToDense(sparseIndices, sparseValues, 
            outputShape, defaultValue)

Parameters: This function accepts four parameters which are illustrated below:

sparseIndices: It is a 0-D, 1-D, or 2-D Tensor of data type int32. Here sparseValues[i] get placed at sparseIndices[i], where i is the indices values.
sparseValues: It is a 0-D or 1-D Tensor of values which corresponds to each row of sparseIndices.
outputShape: It is the shape of the converted dense output tensor.
defaultValue: It is the value to set for indices not specified in sparseIndices. Its default value is zero. It is an optional parameter.

Example 1:

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"
 
// Initializing indices and values
const indices = tf.tensor1d([4, 3, 2, 1, 0], 'int32');
const values = tf.tensor1d([1111, 111, 11, 1, 0.1], 'float32');
 
// Specifying shape for the output dense
const shape = [7];
 
// Getting the Dense representation for the above
// sparse representation
tf.sparseToDense(indices, values, shape).print();

Output:

 Tensor
   [0.1, 1, 11, 111, 1111, 0, 0]

Example 2:

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"
 
// Initializing indices and values
const indices = tf.tensor1d([1, 2, 3], 'int32');
const values = tf.tensor1d([10, 20, 30], 'float32');
 
// Getting the Dense representation for the above
// sparse representation along with shape of [6]
// and default value of 55
tf.sparseToDense(indices, values, [6], 55).print();