By proactively designing it into the event’s schedule, the Rice University Oil & Gas HPC Conference promotes networking – an element of extremely significant value that derives from its workshop roots a decade ago.
The combination of networking and presentations at last year’s event stimulated development of a second poster, “Mitigating Disasters with Deep Learning from Twitter?”, that Univa will present at the 2017 Conference. Although I had already started exploring the value of data from social-networking services like Twitter in various disaster scenarios, my emphasis had been on use of graph analytics. Soon after the 2016 Conference, I completely re-focused my efforts around Machine Learning. In hindsight, I’m really glad that I did re-focus, as Machine Learning has exploded virally over the past year.
Regardless of the disaster scenario, the basic idea of our 2017 poster is to make use of Machine Learning, or more precisely Deep Learning, to classify tweets – a problem that bears a lot of similarity to separating ‘ham’ from ‘spam’ in filtering your email, for example. Our poster emphasizes the need to better connect earthquakes and tsunamis for improved tsunami prediction and communications timelines (e.g., advisories and warnings). In a completely different example, that is even more on message for the RiceU Conference, the case of oil spills receives consideration.
Whether it’s tsunamis or oil spills, the introduction of 6V social networking data has the potential to transform existing, high-stakes problems into ones appropriate for Big Data Analytics – and especially, it appears Machine Learning. The combination of Apache Spark (for Machine Learning) and Univa Grid Engine via Univa Universal Resource Broker is allowing us to obtain in-memory results in scaled-out clusters alongside typical HPC workloads. Our preliminary results appear encouraging in terms of separating geophysically relevant ‘ham’ from ‘spam’ based on the trained models we’ve developed.
If you’re attending the Rice University Oil & Gas HPC Conference, we look forward to seeing you there – you’ll find us taking in talks and posters, as well as in the exhibits area. If you can’t make it to the event, please check back with us here at Univa as we’ll soon share a pointer to our poster on the Big Data of disasters.