Kalimdor.js introduced a new Machine Learning library that works with TypeScript and runs in the browser. How is Kalimdor.js different from other solutions?
Machine learning in the browser based on TypeScript and Node.js – this is possible with Kalimdor.js. The idea behind it is by no means new. Kalimdor is said to be similar to ScikitLearn and even have some of the APIs transferred to TypeScript. This is explained by Jason Shin, developer of Kalimdor, in the media blog post introducing the library.
Machine learning in the browser, explains Shin, is useful in a variety of applications. With Kalimdor.js, for example, it would be possible to have a recommendation system for a music app run locally on the device if no internet connection is available. This does not always work with other solutions. For the implementation of various machine learning functions, Kalimdor supports both supervised and unsupervised machine learning models, including Decision Trees, Random Forests, PCA, KMEans, and SVM. Kalimdor can be used in combination with other machine learning libraries, such as TensorFlow.js.
Kalimdor.js: Machine Learning in the Browser
As you can read on the GitHub page for the project, the introduction to machine learning with Kalimdor.js should also be easier than with other tools. For example, the library includes a number of APIs that can be consistently used across all models to make certain basic functions accessible. This includes training to train a model, for inference, to store the state of a model, and to load models from checkpoints.
Kalimdor.js can be installed via yarn and npm:
If the name of the library seems to be familiar to one or the other: World of Warcraft actually served as a template, as the above-mentioned blog post by Shim shows.