Deeplearn.js providing magnetic features to your browser through Machine Learning

Oct. 15, 2017, 4:03 a.m. By: Vishakha Jha

Deeplearn.js

Can a machine really learn and think...can it really answer all my questions from spelling correction to voice prediction or even in the financial aspect that whether it's a promising stock market or otherwise?

The answer is YES! But if a machine can; so does human, the gulf of the difference comes while comparing the efficiency and the response time which is our basic purpose for an entirely automatized world where everything works on tick of a clock.

Now the next question comes if our machines can already take decisions for us, how we can make it more efficient and interactive? The answer is deeplearn.js which is an initiative from Google Brain PAIR for the browser to redesign the human interactions with ML. You can use the library for everything from education, to model understanding, to art projects and much more.

deeplearn.js is an open-source Web-GL JavaScript library that is browser-based and tends to bring performant machine learning building blocks to the web. It allows you to upskill neural networks in a browser or even execute pre-trained models in inference mode without installations and backend.

There are many reasons due to which we implement ML into our browsers. One of the reasons is that it provides an interactive platform on the client-side which can be used for rapid prototyping and visualization. Along with this although the web ML libraries have been in existence for many years but there are restrictions imposed on the speed of Javascript whereas deeplearn.js has the ability to offer a significant improvement in speed as it considers WebGL for performing computations on GPU. The library can be used with plain JavaScript as well.

It provides two APIs NumPy as for an immediate execution model and another as TensorFlow API for a deferred execution model mirroring. "We have also implemented versions of some of the most commonly used TensorFlow operations. With the release of Deeplearn.js, we will be providing tools to export weights from TensorFlow checkpoints, which will allow authors to import them into web pages for Deeplearn.js inference."

Demos for Deeplearn.js can be viewed on as of the project's homepage. The deeplearn.js provides us a wider scope to implement a great number of features and to have a whole new experience with more ease and flexibility. It has joined a number of projects to showcase the ability that it possess and provide a whole new range of tools.

More Information: GitHub