Motivated by the shortcomings of traditional data challenges, we have developed a unique concept and platform, called Rapid Analytics and Model Prototyping (RAMP), based on modularization and code submission. Open code submission allows participants to build on each other’s ideas, provides the organizers with a fully functioning prototype, and makes it possible to build complex machine learning workflows while keeping the contributions simple. We will start this presentation by describing the context and motivation, the guiding design principles, and some of the technical details (front and backend) of the platform. We will then walk you through some of the most interesting workflows and applications (e.g., anomaly detection in particle physics detectors, classifying molecular spectra for safe drug administration, spatio-temporal time series prediction in climate science). In the last third of the talk we will present a preliminary analysis of the RAMPs that touches on both the technical (machine learning) aspects of the tool and on the management of crowdsourcing data analytics.
More information:
https://drive.google.com/open?id=0BzwKr6zuOkdRNmQ0Q3djMTBzY2s