Projects

The current projects in the lab include:

  • Pliny, a DARPA-funded project aimed at using machine learning methods to mine the world’s code in order to power the next generation of program synthesis, debugging, and repair tools. One of the tools developed by the Pliny project is Bayou.
  • Houdini, a framework for transfer learning based on program synthesis.
  • PIRL, a program synthesis framework for learning interpretable and verifiable policies from reinforcements.
  • NeuroSynth, a family of neurally directed solvers for syntax-guided program synthesis.
  • Robosynth/TMKit, a framework, combining program synthesis and sampling-based methods, for integrated task and motion planning in robotics.
  • PlinyCompute, a high-performance distributed compute platform for the development of data intensive tools and libraries. PlinyCompute is meant to fill the gap between high performance tools such as MPI (which are difficult to use and lack direct support for Big Data) and dataflow platforms such as Spark (which have made many concessions to usability that affect performance, such as reliance of the Java Virtual machine).
  • Applied machine learning, in diverse domains such as oil and gas and biomedical informatics.
  • SimSQL, a project aimed at developing tools to support stochastic analytics. Data almost always have some uncertainty: measurement errors, missing values, etc. It is very natural to characterize uncertainty about data using a probability distribution. SimSQL aims to provide system support for uncertainty characterized in this way, with the goal of scaling to the very largest data sets.

Comments are closed.