Introduction To TensorFlow.js

TensorFlow.js is the popular library of JavaScript that helps us add machine learning functions to web applications. Tensor is the datatype which is used in the TensorFlow.

Now, let us understand the TensorFlow.js and its components.

What is TensorFlow.js?

TensorFlow.js is the JavaScript library that allows us to train, and deploy machine learning models in the browser and on NodeJS. It is open-source and was developed by Google.

• Pre-trained Models: It uses pre-trained models for tasks like text recognition, object detection, etc.
• Cross-Platform Compatibility: It can work on both the NodeJS environment as well as on the web browser.
• Custom Model Training: With less setup, it can train custom models using JavaScript.
• Client-Side Machine Learning: Without severe dependency, it can perform ML computation in the browser

Output

 [1, 2, 3, 4]

In this example

• We will import TensorFlow.js to use its functionalities.
• tf.tensor([1,2,3,4]) creates a tensor with the values [1,2,3,4].
• tensor.print() print the output of the tensor values.

Installing TensorFlow.js

To use TensorFlow.js in a web application, include it using a script tag:

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>

For NodeJS, install the package via npm:

npm install @tensorflow/tfjs

What is a Tensor?

Tensor is the data structure in the TensorFlow.js which represents the multi-dimensional arrays.

• Rank: The number of dimensions in a tensor (scalar, vector, matrix, or higher-dimensional tensors).
• Shape: The size of each dimension in a tensor.
• Data Type: The type of values stored in a tensor (e.g., float32, int32).

Creating a Tensor

We can use the tf.tensor

Output

[1,2,3,4]

• tf.tensor(): Creates a tensor from the provided array. This works in TensorFlow.js and is used to define tensors with the given data.
• tensor.print(): Prints the contents of the tensor to the console in a readable format.

Specifying Shape and Data Type

Output

 [[1, 2], [3, 4]]

• tf.tensor2d(): This function is used to create a 2D tensor (matrix) in TensorFlow.js. It requires three arguments
• matrix.print(): The print() method is called on the tensor object. This method prints the tensor's values to the console in a human-readable format.

Models in TensorFlow.js

TensorFlow.js supports two types of models

Sequential Model

A linear stack of layers, useful for simple neural networks.

const model = tf.sequential();
model.add(tf.layers.dense({ units: 10, inputShape: [5], activation: 'relu' }));
model.add(tf.layers.dense({ units: 1 }));

Functional Model

Functional Models are used for the complex architectures like multi-input and multi-output models.

const input = tf.input({ shape: [10] });
const denseLayer = tf.layers.dense({ units: 5, activation: 'relu' }).apply(input);
const output = tf.layers.dense({ units: 1 }).apply(denseLayer);
const model = tf.model({ inputs: input, outputs: output });

Operations in TensorFlow.js

To manipulate tensors, tensorflow.js provides various mathematical operations.

Addition

Output

[5, 7, 9]

Multiplication

Output

[6, 12, 18]

Dot Product

Output

11

Use Cases of TensorFlow.js

• Real-time Image Recognition: TensorFlow.js allows real-time image recognition in the browser using pre-trained models like MobileNet.
• Speech Recognition: TensorFlow.js can process and analyze audio data for various speech-based applications.
• Natural Language Processing (NLP): TensorFlow.js is also powerful for NLP tasks such as sentiment analysis, text classification, translation, and more.
• Reinforcement Learning: With TensorFlow.js, you can enable users to interact with web applications using gestures instead of traditional input devices like keyboards and mice.