91短视频 (91短视频) senior Alex Bender presented in the Student Research Competition at this year鈥檚 Association for Computing Machinery鈥檚 Technical Symposium on Computer Science Education, popularly known as SIGCSE. The annual event, held this year in Memphis March 2-5, draws many of the top scientists in the field.
Bender, a senior major from Sarasota, Florida, presented a research poster titled 鈥淎n Evaluation of Cluster 3.0 as a General Tool for Principal Component Analysis.鈥� Professor , who advised Bender on the project, says that he is only one of two students to have work accepted to the prestigious venue since she鈥檚 been at 91短视频. The other student was Aaron Springer, a 2013 graduate who is currently earning his doctorate at the University of Santa Cruz.
鈥淎aron鈥檚 work was completed during a summer Research Experience for Undergraduates, so this could be the first work that was actually done at 91短视频 to be presented at SIGCSE,鈥� Weikle said.
Developed criteria for choosing software
Principal Component Analysis (PCA), Bender says, is a way for a researcher to take a very large data set and reduce it to a much smaller set of data points that represent the most important parts of the data. That, in turn, makes understanding and drawing analyses or conclusions from the data much simpler.
鈥淢y research led me to discover that in many computer science or computer architecture research papers where PCA is utilized, the reasoning and explanation for how or why PCA was used were inconsistent at best and nonexistent at worst,鈥� Bender says. 鈥淲e also knew that PCA had uses鈥攁nd was being used鈥攐utside the pure mathematical or computer science-related fields.鈥�
If a researcher wanted to utilize PCA for research purposes, Bender concluded that 鈥渋t would be useful to create a general set of criteria for choosing software that implemented PCA well.鈥�
Bender developed six key criteria against which to judge potential software. The criteria included whether it was an open-source program; descriptive, clear documentation; data compatibility to ensure correct processing; built-in assumption testing; built-in graphical tools to allow for clear depictions; and the ability to sufficiently process large amounts of data.
In his tests with the software program Cluster 3.0, Bender found that it passed four of the six criteria and had 鈥渞easonable ways to cope鈥� with the shortcomings in the other two criteria, thus deeming the software viable for PCA implementation in scientific experiments.
Bender鈥檚 presentation at SIGSCE was a big step from last semester鈥檚 , where he鈥檇 presented the PCA-related portion of a research project that he鈥檇 completed with Samuel Miller on the ripening process of noni fruit.
鈥淚t being my first time talking about research聽at all, much less in front of a crowd, I felt reasonably good about our presentation,鈥� Bender says. 鈥淚t definitely helped prepare me to talk to the judges and interested participants at SIGCSE.鈥�
Research began with independent study
Bender鈥檚 intensive work with PCA began during his junior year, when he took an independent study with Weikle based around PCA鈥檚 theory and underlying mathematics along with some computer architecture. After the study ended, Bender says they decided to continue the research to work toward a tangible result. That extended into the summer, when Bender was invited to attend 91短视频鈥檚 Kairos Place, an annual event where faculty, staff and graduate students can work on research, writing and creative projects.
鈥淚t was a fascinating experience seeing all the different avenues of research, and different people鈥檚 processes in their work, and seeing interactions between scholars and teachers that I may otherwise never have gotten to witness,鈥� Bender says. 鈥淜airos was the point of clarification, where Dee and I worked out more clearly what the goal of our research would be.鈥�
After graduation, Bender says he plans to work for a few years in software development to gain experience and refine his research interests before pursuing graduate school.