Research Projects

My research interests include the design and analysis of algorithms, computational geometry, data mining and machine learning, and the application of such algorithms to interdisciplinary data.

The main research projects I have worked on are described in detail below, with associated publications, grants, and patents.

Algorithmic Fairness
Understanding Motion
The Dark Reaction Project
Beyond the Red Pen
Indoor Location for Google Maps (Google [x])

Algorithmic Fairness

I'm interested in understanding how data mining and machine learning algorithms could be unfair and what can be done to prevent that. I'm working with Suresh Venkatasubramanian and Carlos Scheidegger on examining this question in the context of the legal notion of disparate impact.

For more information, see fairness.haverford.edu.

Papers

Michael Feldman, Sorelle A. Friedler, John Moeller, Carlos Scheidegger, and Suresh Venkatasubramanian. Certifying and Removing Disparate Impact. Proceedings of the 21st ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2015. [PDF]

Understanding Motion

I'm interested in understanding motion from a computational geometry and information theoretic point of view; creating frameworks and analyses that are theoretically sound and yet practically relevant. I'm working with Dianna Xu and Betul Atalay on related ideas for practical frameworks for kinetic data. The preliminary version of this work was presented at the 2013 Fall Workshop on Computational Geometry.

Thesis

Sorelle A. Friedler. Geometric Algorithms for Objects in Motion. Dissertation committee: Prof. David Mount (chair), Prof. William Gasarch, Prof. Samir Khuller, Prof. Steven Selden, Prof. Amitabh Varshney. Defense date: July 30, 2010. [PDF] [presentation]

Papers

Sorelle A. Friedler and David M. Mount. A Sensor-Based Framework for Kinetic Data Compression. Computational Geometry: Theory and Applications, 2014. (doi: 10.1016/j.comgeo.2014.09.002) [PDF (preprint) | link]

Sorelle A. Friedler and David M. Mount. Spatio-temporal range searching over compressed kinetic sensor data. In Proc. of the European Symposium on Algorithms (ESA), pages 386-397, 2010. [PDF (preprint) | link] [TR]
     2nd Workshop on Massive Data Algorithmics, 2010 [PDF]
     Fall Workshop on Computational Geometry, 2009 [PDF]

Sorelle A. Friedler and David M. Mount. Realistic compression of kinetic sensor data. Technical Report CS-TR-4959, University of Maryland, College Park, 2010. [PDF | TR]

Sorelle A. Friedler and David M. Mount. Approximation algorithm for the kinetic robust k-center problem. Computational Geometry: Theory and Applications, 2010. (doi: 10.1016/j.comgeo.2010.01.001). [PDF (preprint) | link]

Sorelle A. Friedler and David M. Mount. Compressing kinetic data from sensor networks. In Proc. of the 5th International Workshop on Algorithmic Aspects of Wireless Sensor Networks (AlgoSensors), pages 191 - 202, 2009. [PDF (preprint) | link] [TR]

In Submission

F. Betul Atalay, Sorelle A. Friedler, and Dianna Xu. Convex Hull for Probabilistic Points. [PDF]

The Dark Reaction Project

I'm working with Chemists Josh Schrier and Alex Norquist as well as computer science student Paul Raccuglia '14 on an applied machine learning project. We are collecting experimental data from materials chemistry experiments and using the outcomes to predict what future experiments might be successful.
Our model currently achieves a 93% success rate when choosing between four possible outcome states. We will continue this work by increasing the size and reach of the data set and building a recommendation system that will suggest future potential successful experiments.

Grants

NSF DMR-1307801 (2013 - 2016): The Dark Reaction Project: a machine learning approach to materials discovery. Joshua Schrier, Alexander Norquist, and Sorelle Friedler. $299,998.

Beyond the Red Pen


I am interested in the ways in which technology can aid teachers, and thus aid students, by making the tedious parts of teaching disappear while simultaneously granting teachers more insight into students’ progress.

Previously, I've worked with Ben Shneiderman, Yee Lin Tan, and Nir J. Peer to visualize grades so that teachers can identify individual student strengths and weaknesses. Interviews with sixteen expert teachers across multiple disciplines indicated that the software was successful at providing teachers a useful tool to understand their students' progress and encourage reflective practice. The previous project page, including links to the sourcecode, help videos, and journal paper can be found here.

Papers

Sorelle A. Friedler, Yee Lin Tan, Nir J. Peer, and Ben Shneiderman. Enabling teachers to explore grade patterns to identify individual needs and promote fairer student assessment. Computers & Education, 51(4):1467-1485, December 2008. [PDF (preprint) | link]

Indoor Location for Google Maps (Google [x])

As part of the indoor location team within Google X, I helped to use and develop applied machine learning techniques to implement an indoor location determination system running within Google Maps for Mobile on Android. The blog post announcing our launch can be found here.

When trying to locate a person inside a building using only a cell phone, GPS cannot be relied upon and so other phone sensors must be used instead. These sensors were not designed to be used for location purposes and measurements collected by them tend to be very noisy. These issues combine to create a hard machine learning problem that can be solved by making use of probabilistic graphical models and sensor fusion to locate a person indoors.

Patents

Mohammed Waleed Kadous, Isaac Richard Taylor, Cedric Dupont, Brian Patrick Williams, Sorelle Alaina Friedler. Permissions based on wireless network data. US 20130244684 A1. Publication date: Sep. 19, 2013.

Sorelle Alaina Friedler, Mohammed Waleed Kadous, Andrew Lookingbill. Position indication controls for device locations. US 20130131973 A1 (also WO 2013078125 A1). Publication date: May 23, 2013.