Past Projects

Our team members have previously contributed to a number of public research and development projects. We include here a summary of some of the most notable.
 

Data driven discovery of models (d3m)

Organized by DARPA, the D3M project is a large-scale cross-institution effort to develop new tools for simplifying and eventually automating the difficult work of building novel machine learning systems. The project assembled a large library of individual modules for performing specific data processing tasks, and combined it with AI systems capable of linking these modules together into complete, coherent pipelines. The goal, ultimately, is to build a system that can receive a description of a novel machine learning problem from a human, and automatically search a massive space of possible methods for solving the given problem before returning the best available solution.

Rapid optimization of commerical knowledge (ROCK)

The ROCK program was an 8 year long public-private partnership initiative that worked with the regional industries of Rockford and northern Illinois to improve the technology utilization of companies working as parts supplies to the armed forces. The project sought to ensure that new techniques, tools and practices developed in academia actually made their way into real-world use by companies that could benefit from them. The ROCK program was later spun off into a new manufacturing research lab called the Quad City Manufacturing Laboratory to help increase that region’s state of manufacturing technology on an ongoing basis.

Aggregative contingent estimation (ACE)

The ACE project was an IARPA-sponsored effort to use machine learning technology to maximize the impact and utilization of human expertise. In many domains, human knowledge and judgement remain the most valuable tools for making sense of complex data. The ACE program sought to build and improve mathematical models for combining and weighting feedback from many human sources, and extrapolating from that pooled data to produce more complete and accurate predictions.