BrainJS is a library that makes it easy to create neural networks and then train them based on input/output data. Since training consumes a lot of resources, it is preferable to run the library in a Node.js environment, although the CDN browser version can also be downloaded directly to the web page. There is a tiny demo on their site that can be trained to detect color contrast.
An educational web application that lets you play with neural networks and explore their various components. It has a nice interface that allows you to control the input data, the number of neurons, choose the algorithm, and other indicators that will be reflected in the final result. In addition, there is a lot of other useful information: the application is open source and uses a special machine learning library written in the TypeScript language. The fact that it has good documentation also serves as a plus.
FlappyLearning is a JavaScript project that contains about 800 lines of code, allows you to create a machine learning library and implement it in a fun way that teaches you to play Flappy Bird , like a virtuoso. The artificial intelligence technique used in this library is called Neuroevolution and applies algorithms based on nervous systems found in nature, dynamically learning from the success or failure of each iteration. The demo is very easy to run - just open index.html in your browser.
Probably the most actively supported project on this list, Synaptic is a Node.js and browser library that is architecture-independent, allowing developers to build any kind of neural network. It has several built-in architectures, which allows you to quickly test and compare different machine learning algorithms. In addition, Synaptic is a well-written introduction to neural networks, a number of hands-on demonstrations, and many other great tutorials that expose how machine learning works.
Land Lines is an interesting Chrome web experiment that finds satellite images of the Earth that look like doodles made by the user. The app makes no server calls: it runs entirely in the browser and has excellent performance, even on mobile devices, thanks to its clever use of machine learning and WebGL. You can check out the source code on GitHub or read the full study here.
Although it is no longer actively supported, ConvNetJS is one of the most advanced deep learning libraries for JavaScript. Originally developed at Stanford University, ConvNetJS has become quite popular on GitHub, resulting in many features and tutorials. It runs directly in the browser, supports multiple learning methods, and is fairly low-level, making it suitable for people with a lot of neural networking experience.
Thing Translator is a web experiment that allows your phone to recognize real objects and name them in different languages. The app is built entirely with web technologies and uses two machine learning APIs: from Google - Cloud Vision for pattern recognition and Translate API for natural language translation.
A framework for creating artificial intelligence based on reinforcement learning. Unfortunately, the open-source project does not have proper documentation, but one of the demonstrations, an experiment with autopilot in a car, has an excellent description of the different parts that make up the neural network. The library is written in pure JavaScript and made using modern tools such as webpack and babel.
Another library that allows us to configure and train neural networks using only JavaScript. It is very easy to install, both in Node.js and on the client side, and has a clean API that will be easy for developers of any skill set. The library provides many examples that implement popular algorithms to help you understand the basic principles of machine learning.
DeepForge is a convenient development environment for working with deep learning. It allows you to design neural networks using a simple interface, supports training models on remote machines, and has built-in version control. The project runs in a browser and is based on Node.js and MongoDB, making the installation process very familiar to most web developers.
Useful links: