Stardog 5 RC1 includes cutting-edge machine learning and a new monitoring system.
First, let’s do some housekeeping about our release process for Stardog 5:
With today’s RC1 release, Stardog 5.0 is feature complete. The only changes, if any, before 5.0 final will be bug fixes. Which means the next two items are happy news; if they hadn’t gotten into the RC1 release window we’d all be sad.
Pedro Oliveira blogged here recently (Learning to Predict) about our work to deeply integrate machine learning into Stardog—in this first iteration by integrating Vowpal Wabbit into Stardog. In particular we wanted to make a range of learning services available to Stardog Knowledge Graph users, taking advantage of the potent combination of better data and better algorithms. And we wanted this integration to be as seamless as possible, including model training, ideally at the level of graph query evaluation, i.e., integrated with the ordinary yet powerful machinery of SPARQL: reads and writes of real and virtual graphs.
We’re happy to announce that machine learning is included in the RC1 release and will be in Stardog 5.0 final. There’s a separate post—The Benefits of Learning—which focuses on the business benefits of machine learning in an Enterprise Knowledge Graph.
We also managed to sneak a new monitoring framework into RC1. Based on Metrics, the monitoring approach in Stardog is more consistent with best practices for enterprise platforms.
Stardog 5 exposes far more of the system’s internals in the monitoring output, which will help all of us—you and us—make better decisions going forward. Stardog 5 now reports
A subset of this info will be available by default in the CLI, and we’ve removed the monitoring visualization from the Web Console completely. Consistent with best practice, we now recommend customers use a dedicated monitoring tool like Graphite or Ganglia or some other enterprise standard.
Download Stardog today to start your free 30-day evaluation.