Federal Government Challenges

Federal Government Challenges

 

EPA Cyanobacteria Modeling Challenge

The goal of this project was to predict the locations and times that the cyanobacteria bloom to occur. Algae are natural components of marine and fresh water flora performing many roles that are vital for the health of ecosystems. However, excessive growth of algae becomes a nuisance to users of water bodies for recreation activities and to drinking water providers. Moreover, “blooms” are formed, the risk of toxin contamination of surface waters increases especially for some species of algae with the ability to produce toxins and other noxious chemicals (problem statement). A three week, $15,000 prize purse competition fielded 30 unique code submitters and 460 submissions. The winning algorithm predicts the cyanobacteria concentrations with an accuracy of an order of magnitude. Subsequently, a user app for mobile phones was developed for practitioners’ use (2014). 

EPA ToxCast Challenge

The goal of this challenge was to develop a model based on data provided by the EPA to quantitatively predict a chemical’s systemic lowest effect level (LEL) in a traditional animal toxicity study. The systemic Lowest Effect Level or LEL is the lowest dose that shows adverse effects in these animal toxicity tests. The LEL is then conservatively adjusted in different ways by regulators to derive a value that can be used by EPA to set exposure limits that are expected to be tolerated by the majority of the population (problem statement). Entrants to this challenge had to develop models using data from high-throughput in vitro assays, chemical properties, and chemical structural descriptors to quantitatively predict a chemical’s systemic LEL. 47 Competitors submitted 783 solutions for a $10,000 prize purse. The result yielded an improvement by 20% in the accuracy of prediction above the current state of art. For more information, please visit the project website (2014).

Tech Challenge for Atrocity Prevention

USAID and Humanity United tasked competitors to develop an algorithm to predict where and when atrocities will happen in the near future (within the next 30 days) based on the atrocity pre-history (Political Instability Task Force Data) and sociopolitical activities (Global Database of Events, Language, and Tone). The problem statement is available here. In all, 93 competitors produced 592 submissions over a three-week period for a $25,000 prize purse. The winning solution developed a reliable algorithm that can show the heat-map of atrocity risks worldwide in real-time. The prediction power is 60% better than the one of predictions, based just on atrocity frequency in the location. For more information, please visit the project website (2013).

U.S. Patent Trademark Office

The Image Processing Algorithm for Patent Data Challenge tasked competitors to develop algorithms that automatically detect figures and part labels in U.S. patent drawing pages. The challenge drew 232 teams of two, of which 70 teams (30%) submitted solutions. Collectively, teams submitted 1,797 solutions that were compiled on the competition servers. Participants reported spending an average of 63 hours developing their solutions, resulting in a total of 5,591 hours of development time. A manually labeled dataset of 306 patents was used for training, online system tests, and evaluation. For the first place system, the harmonic mean of recall and precision (f-measure) was 88.57% for figure region detection, 78.81% for figure regions with correctly recognized figure titles, and 70.98% for part label detection and character recognition (2011).