Research <- Sean Forman <- You Are Here
|
I have left my academic position, so this represents my research interests while in that position. I am now mostly working as the founder and president of Sports Reference, Inc..
Protein Folding
Protein Folding involves translating the protein's sequence of amino acids into the 3-D structure of every atom in a protein. I have worked for several years with Alberto Segre on the protein folder, HOPS. HOPS defines a search space of potential conformations using a full protein representation along with a set of discrete torsion angles. It then travel through the resulting search tree avoiding steric clashes and looking for the protein's minimal energy state. This approach is analogous to computer chess, which uses an array of secondary search techniques layered above a single underlying search paradigm.
I am looking to re-start this work for the following year or two beginning Summer 2003.
Automatic Congressional Redistricting
I have also studied genetic algorithms and have written a program which attempts to draw compact United States congressional districts. This technique adapts a genetic algorithm designed for traveling salesman problems. Working with a student, I have implemented a hybrid system that uses the genetic algorithm to produce a rough cut solution and then a swapping technique to more fully balance district populations. This work has led to a pair of submissions to the Genetic and Evolutional Computation Conference (GECCO).
Possible work in this area includes a departure for the TSP-based technique used currently and compare that with a grouping genetic algorithm.
TSP Generator
This web application will accept a list of up to 30 U.S. cities, and then compute the inter-city distances, call on the census bureau's web server to create a map, and then compute approximate solutions to the resulting Traveling Salesman Problem. For small problems it will use a branch-and-bound complete solver to find an optimal tour. This is not an advanced TSP solver, but an educational tool. It can be a useful resource for instructors in undergraduate contemporary mathematics courses and operations research courses.
Web Design
I have experience designing web sites and am interested in pursuing work in both teaching web design and designing web sites for other educational purposes. I designed and implemented a pre-calculus testing service which St. Joseph's used to test 600 incoming freshman students. I also have implemented weblogging technologies into my course instruction. This enables easy updates of my course web sites and are something other professors might find useful.
In Spring 2003, I developed a new course, Internet Applications Development, for our graduate students. I intend to continue teaching this course each year.
Nagging and Satisfiability
Nagging is a novel parallel computing technique used to solve combinatorial search problems. Alberto Segre is almost entirely responsible for developing the nagging infrastructure. Nagging takes advantage of the random nature of combinatorial optimizaton by, for example, permuting the order in which nodes are searched within a search tree. It is a general purpose optimization tool and we have implemented it for alpha-beta minimax search, the traveling salesman problem and other combinatorial search problems.
I have written a 3-SAT (satisfiability) solver for nagging. Solving 3-SAT involves either finding a valid solution or proving no solution exists for a conjunction of three-variable literal clauses. For each of these problems, nagging displays a varying level of super-linear speedup. A general paper on nagging has been given on this topic. I worked with two graduate CS students on incorporating more sophisticated 3-SAT solver techniques into our current parallel solver. This is work that I am no longer continuing.
|