Happy July 4th! This week was short but productive. I ran all the negative tests for geoHash, which took an exceptionally long time. Brian wrote two versions of geoHash. The first version lacks the ability to run the graphical interface but can run on more of the computers in the office, including the cluster. Unfortunately, however, this version of geoHash cannot run negative tests. Consequently, I had to run all of the negative tests on Zeus, one of the computers in the office. In total, it took 5 days, running day and night to finish the negative tests. Fortunately, I first automated the entire process so that I could just let it run.
While the tests ran, I started on a new assignment. Brian gave me a paper to read and asked me to prepare a presentation once I was done. The paper, Protein Structure Alignment Using Dynamic Programming and Interactive Improvement, by Tatsuya Akutsu described three methods for aligning 3-D protein structures. They all consist of a similar core algorithm which takes two triples of points, finds the translation with the smallest distance between the two triples, then uses a formula to determine how good the match is. The difference between the three comes in the manner in which the pairs of points are chosen. The first takes all possible combinations. The second chooses a random group of points then takes combinations from that group. The third tests all triples, but performs the transformation and matching algorithms only if the root mean square distance is small enough. Once I have reported on the paper, Brian wants me to implement some sort of 3-D matching algorithm for protein structures. Hopefully the paper will give me some ideas on how to do this.