Mark A Coletti

Mark Coletti

Mark A Coletti

Staff Scientist

Bio

As a staff scientist at the Oak Ridge National Laboratory, I am actively performing research in the areas of geoinformatics, machine learning, and evolutionary computation.  My current principal research focus is on using machine learning to identify areas of human habitation in satellite imagery. I am also involved in research related to Volunteered Geographic Information (VGI) and agent-based modeling.

Previously, I was the Chair of the Penn State Postdoctoral Society, and as such was responsible for organizing career enhancement, personal improvement, and social activities for over 460 postdoctoral scholars.  Also, I worked at George Mason University's Center for Social Complexity and Evolutionary Computation Laboratory where I developed an evolutionary computation C++ toolkit; a biologically inspired cognitive model for a DARPA Grand Challenge; a Joint Improvised Explosive Device Defeat Organization related multiagent simulation; an Office of Naval Research  Multidisciplinary University Research Initiative Office sponsored massive multiagent simulation of pastoral and farming behavior in eastern Africa; and a geospatial extension, GeoMason, for the multi-agent simulation toolkit MASON.  

Earlier in my career I worked as a senior software engineer in the Washington, DC, area on projects for the National Oceanic and Atmospheric Administration, Federal Highway Administration, U. S. Army's Materiel Command, the U. S. Army Topographic Engineering Center, and the United States Geological Survey.  These projects included an expert system to correct human sourced sea surface meteorological data, an expert system for validating materiel purchases, a topographic visualization system, a road surface wear calculator, and a toolkit for spatial data format conversion.   

Awards

July 2009, Best Graduate Student Workshop Paper at GECCO

Projects

Settlement Mapper Tool

I work for the Geographic Information Science and Technology group on the Settlement Mapper Tool, which automatically identifies inhabited areas in satellite images using computer vision techniques.  My responsibilities include refactoring CUDA GPU C/C++ code and python scripts, and improving the use of machine learning and vegetation indices for identifying inhabited structures in satellite imagery.  To that end, I have developed a tool that does radiometric scaling of DigiGlobe satellite imagery in about one sixth the time of equivalent commercial software. I also developed open source software, GPU Manager, that manages multiple Nvidia GPUs among multiple users.  Currently, I am exploring the use of evolutionary algorithms to optimize deep learner architectures and hyper-parameters using our HPC resources, Titan and Summit.

 

Eagle-I

The Eagle-I project provides a dashboard depicting the health of the United States power grid and fossil fuel pipelines.  I am the principal architect of a new design for the software responsible for gathering power outage information for the over 300 power utilities in the US.  This new system has cut in half the number of scripts needed, and reduced the collective polling time from over nine minutes to around five.  Moreover, this new approach allows for easier maintenance; for example, it took two days to compensate for changes a utility made on its web site with the legacy system, and about 30 minutes to make the same changes with the new.

Publications

Books

Mark Coletti: The GeoMason Cookbook. 01/2013;

Book Chapters

William G. Kennedy, Chenna Reddy Cotla, Tim Gulden, Mark Coletti, Claudio Cioffi-Revilla: Towards Validating a Model of Households and Societies in East Africa. Advances in Computational Social Science, 01/2014. Pages 315-328; ISBN 978-4-431-54846-1

Journal Publications

Coletti, M., Hultquist, C., Kennedy, W. G., & Cervone, G. (2017). Validating Safecast data by comparisons to a US Department of Energy Fukushima Prefecture aerial survey. Journal of environmental radioactivity, 171, 9-20.

William G Kennedy, Atesmachew B Hailegiorgis, Mark Rouleau, Jeffrey K Bassett, Mark Coletti, Gabriel C Balan, Tim Gulden: An Agent-Based Model of Conflict in East Africa And the Effect of Watering Holes.

Atesmachew B Hailegiorgis, William G Kennedy, Mark Rouleau, Jeffrey K Bassett, Mark Coletti, Gabriel C Balan, Tim Gulden: An Agent Based Model of Climate Change and Conflict among Pastoralists in East Africa.

Alexei V. Samsonovich, Giorgio A. Ascoli, Kenneth A. De Jong, Mark Coletti: Integrated Hybrid Cognitive Architecture for a Virtual Roboscout.

Conference Proceedings

Mark Coletti: The Effects of Training Set Size and Keeping Rules on the Emergent Selection Pressure of Learnable Evolution Model. Genetic and Evolutionary Computing Conference Proceedings; 01/2012

Mark Coletti, Guido Cervone: Analysis of Emergent Selection Pressure in Evolutionary Algorithm and Machine Learner Offspring Filtering Hybrids. Swarm, Evolutionary, and Memetic Computing. Springer Berlin Heidelberg, 2012. Pages 721-728;  ISBN 978-3-642-35379-6

William G Kennedy, M Rouleau, Jeffrey K Bassett, Mark Coletti, Gabriel Catalin Balan, Tim Gulden, Claudio Cioffi-Revilla: MASON HerderLand: Origins of Conflict in East Africa. Talk presented at the annual meeting of the Computational Social Science Society; 01/2010

William G Kennedy, Atesmachew B Hailegiorgis, Tim Gulden, Jeffrey K Bassett, Mark Coletti, Gabriel Catalin Balan, Meghan Clark, Claudio Cioffi-Revilla: An Agent-Based Model of Conflict in East Africa and the Effect of the Privatization of Land. Proceedings of the 3rd World Congress on Social Simulation; 01/2010

Michael Q. Kalish, Alexei V. Samsonovich, Mark Coletti, Kenneth A. De Jong: Assessing the role of metacognition in GMU BICA. Biologically Inspired Cognitive Architectures 2010 - Proceedings of the First Annual Meeting of the BICA Society, Washington, DC, USA, November 13-14, 2010; 01/2010

M Rouleau, Mark Coletti, J K Bassett, A B Hailegiorgis, T Gulden, W G Kennedy: Conflict in Complex Socio-Natural Systems: Using Agent-Based Modeling to Understand the Behavioral Roots of Social Unrest within the Mandera Triangle. Human Behavior-Computational Modeling and Interoperability Conference; 01/2009

Jeffrey K. Bassett, Mark Coletti, Kenneth A. De Jong: The relationship between evolvability and bloat. Genetic and Evolutionary Computation Conference, GECCO 2009, Proceedings, Montreal, Québec, Canada, July 8-12, 2009; 01/2009

Mark Coletti: Learnable evolution model performance impaired by binary tournament survival selection. Genetic and Evolutionary Computation Conference, GECCO 2009, Proceedings, Montreal, Québec, Canada, July 8-12, 2009, Companion Material; 01/2009

Mark Coletti: A preliminary study of learnable evolution methodology implemented with C4.5. Evolutionary Computation, 2002. CEC '02. Proceedings of the 2002 Congress on; 06/2002

Lynn Usery, George Timson, Mark Coletti: A Multidimensional Geographic Feature System. Proceedings of GIScience; 01/2002

Mark Coletti, Thomas D. Lash, Ryszard S. Michalski, Craig Mandsager, Rida E. Moustafa: Comparing Performance of the Learnable Evolution Model and Genetic Algorithms.. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 1999), 13-17 July 1999, Orlando, Florida, USA; 01/1999

Technical Reports

Keith Sullivan, Mark Coletti, Sean Luke: GeoMason: Geospatial Support for MASON. 2010.

Nicholas Payette, Marius Bujorianu, Glen Ropella, Ken Cline, Jeffrey Schank, Matt Miller, Sara Jonsson, Laszlo Gulyas, Richard Legendi, Olaf Bochmann, Lu´ıs de Sousa, Vlasios Voudouris, Daniil Kiose, Przemyslaw Szufel, Steve Saul, John McManus, Vittorio Scarano, Gennaro Cordasco, Chris Hollander, Paul Wiegand, Vera Kazakova, Brian Hrolenok, J. Daniel Rogers, Michael Schader, Sean Luke, Kenneth De Jong, Mark Coletti, Paul Schopf, Claudio Cioffi-Revilla, Keith Sullivan, Khaled Talukder, Ahmed Elmolla, Ermo Wei: Future MASON Directions: Community Recommendations Report of the 2013 MASON NSFWorkshop.

Facilities Used

Titan and Summit

Specialized Equipment

I use the Titan and Summit supercomputer platforms for my evolutionary computation / deep learner research.

User Facility

Contact Information