Brain is an easy library to create neural networks that you can train based on its input/output data. Because the training needs a lot of data resources, although there are CDN links can be directly loaded on the web page through the network. However, it is recommended to use node.js to run the library. Here’s an example, on their website, of being trained to recognize color contrasts.

Educational Web applications let you engage with the world of neural networks and explore their different components. It has a nice UI that allows you to control the input of data, the number of neurons, what algorithm to use, and a lot of other tunable parameters to influence the final result. There’s also a lot to learn from in-app scenarios — the code is open source and uses a definable machine learning library that is typescript-based and well-documented.

FlappyLearning is a JavaScript project with about 800 lines of uncompressed code that creates a machine learning library to implement a fun example of playing Flappy Bird. The artificial intelligence technology used in the library, called Neuroevolution, uses natural neural algorithms to learn dynamically based on the success and failure of each iteration. This example is very easy to run – just open the index.html file in a browser.

Probably the most active project on this list, Synaptic is a Node.js and browser-available library that is an agnostic construct that allows developers to build any type of neural network. It has some constructs that allow it to test and compare different machine learning algorithms more quickly. There is also a good introduction and documentation, some examples of exercises, and a lot of great instructions to understand how machine learning works.

Land Lines is a very interesting Chrome Web attempt to find satellite images of earth. It’s like letting users doodle. The application doesn’t have any server-side requests: it runs entirely in the browser, thanks to clever use of machine learning, and WebGL’s fantastic performance, even on mobile devices. You can find the source on GitHub here or read the whole case.

Although no longer actively maintained, ConvNetJS is the most in-depth JavaScript learning library. Originally developed at Stanford University, ConvNetJS became very popular on GitHub, with many communities starting to drive new features and manuals. It runs directly in the browser, supports multiple learning techniques, and is very rudimentary, which makes it more suitable for people to experience neural networks.

Thing Translator is a web experiment that lets your phone recognize real objects in different languages. The app is powered entirely by Web technology and incorporates Google’s two machine learning apis – Cloud Vision for image recognition and the Translate API for neuro-linguistic translation.

Construct the framework of AI system based on reinforcement learning. Unfortunately, there is no proper documentation for this open source project. But one example, an experiment in self-driving cars, has detailed descriptions of different parts of the neural network. The library is pure JavaScript and uses WebPack and Babel.

Another library that allows you to build and train neural networks using only JavaScript. Very easy to install, requires Nodejs and a client, very easy for developers to use API calls. This library provides a number of examples to implement popular algorithms to help understand the core machine learning principles.

DeepForge is a user-friendly development environment for using deep learning. Allow yourself to design neural networks with some graphical interfaces, support training models for remote machines, and have version control. The project runs in a browser, based on Node.js and MongoDB, and the installation process is familiar to Web developers.

Foley: Machine Learning in Javascript

Excellent blog post by Burank Kanber on the fundamentals of machine learning. Perfect for JavaScript developers to read and learn. If you want to learn more about machine learning, this is a great resource to recommend.