Open-Source ML Server Gets Apache Promotion
The pace of machine learning technology development got another boost this week with the announcement that an open source platform donated last year by Salesforce has been promoted by the Apache Foundation.
Apache PredictionIO, designated a “top-level” project on Tuesday (Oct. 24), aims to democratize machine learning by giving developers a full stack for creating intelligent applications that could be deployed in production “without having to cobble together underlying technologies,” said Simon Chan, founder of Prediction IO who now services as senior director for Salesforce’s AI initiative called Einstein.
PredictionIO was launched around an open source machine-learning server. Its “template gallery” now includes tools ranging from classification and natural language to recommendation and regression engines. Chan noted that the downloadable engines would significantly reduce development time. For example, a recommendation engine that would normally take data scientists months to build could be developed by several engineers “in a couple weeks,” Chan claimed.
Acquired by Salesforce in 2016, PredictionIO is part of the company’s Einstein AI push. Salesforce developers previously used the machine-learning platform to create an “agenda builder” app designed to suggest relevant sessions when attending technical conferences. The application gauges user preferences along with those of other attendees to develop individual conference agendas, Chan explained.
Other applications developed using PredictionIO include tools for predicting college dropout rates and custom predictive models for banking customers.
Since Salesforce( NYSE: CRM) donated an open-source version of PredictionIO to the Apache Foundation last year, Chan said his team has been working with machine learning mentor while building up a user and developer community to push the project from incubation to a full-fledged, “production-ready” Apache project.
When Salesforce acquired PredictionIO in February 2016, the startup was folded into the Salesforce machine learning initiative with the goal of adding intelligence across its cloud-based customer relationship management platform.
The open-source approach promises to address what one recent industry survey identified as slow adoption of machine learning despite a consensus that the technology is central to data analytics and AI efforts. The survey released in September by software giant SAP found that only 7 percent of 3,100 companies polled are currently investing in machine learning development.
The Apache Foundation said PredictionIO could be installed as a full machine-learning stack, bundled with Apache Spark, MLLib, HBase, Spray and ElasticSearch. Among other capabilities, the machine-learning server helps developers build predictive engines for machine learning tasks.
More details and release notes are here.