Sign In

Communications of the ACM

ACM TechNews

Google's Go Language Ventures Into Machine Learning

View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
Artist's represenation of artificial intelligence.

A small but growing number of libraries written entirely in Go are stepping up to show that the programming language can handle machine learning.

Credit: Thinkstock, Inc.

Although existing machine-learning libraries in other coding languages, such as C/C++, have a far larger culture of users, there is a growing interest in having Go toolkits that take advantage of the language's conveniences.

For example, GoLearn, which Google describes as a "batteries included" machine-learning library, is one of the most widely used. Some of GoLearn's interfaces to data handling are implemented in the same way as scikit-learn, a popular Python machine-learning project. Although there is some C++ used for the linear models library, most of the platform is solely based on Go.

Meanwhile, Goml is "Golang machine learning, on the wire," as it includes many models that enable users to learn in an online, reactive manner by passing data to streams held on channels. The project emphasizes its possibilities as a component for other applications, made easier by comprehensive tests, extensive documentation, and clean, expressive, modular source code.

Gorgonia is a machine-learning library written entirely in Go, and it provides the necessary primitives to dynamically build neural networks and assorted machine-learning algorithms.

From InfoWorld
View Full Article


Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA


No entries found