DIMACS

Project Description

For my research project, I'll be working on an information fusion paradigm called Combinatorial Fusion Analysis. I will be using combinatorics and statistics to find the most effective way to combine data. Applications of this paradigm include information retrieval, target tracking, cyber security, virtual screening, DNA sequencing informatics, bioinformatics, and cognitive neuroscience. I will also be analyzing the relationship between various methods of correlation and cognitive diversity to further understand successful information fusion.


Weekly Log

Week 1:
During my first week, I read two papers, Combining Two Visual Cognition Systems Using Confidence Radius and Combinatorial Fusion and Rank-Score Characteristics (RSC) Function and Cognitive Diversity. I researched the statistical methods used in these papers and began compiling a data set to test.
Week 2:
During my second week, I learned python and wrote a script to analyze the data from one of the papers that I read. I began researching a method to generate data from a limited set of data in order to be able to get conclusive results with a limited set of data. I also researched various papers on fusion problems.
Week 3:
During my third week, I explored graphing in Python. I created graphs to visualize my data in an effective way.I also learned about three statistical methods for approximating the correlation in different sets of data. I used these methods to look for a pattern in my data, depending on how correlated each set of data was. I began researching an effective way to use bootstrapping to generate more data for the problem I'm working on.
Week 4:
During my fourth week, I wrote a script using bootstrapping to resample my data and generate more meaningful results. I then found a more effective way, and spent the week editting my script and generating new tables and graphs with the hope of finding a pattern in my data. I then worked on narrowing down each test to find the most efficient way to understand the new data I had generated. After running the new script, the patterns in the data became more evident.
Week 5:
During my fifth week, I analyzed the results of the previous weeks. I ran tests and generated new data and tried to formulate a hypothesis that held in each trial. I explored different methods of correlation. Finally, I further developed my graphs and learned new tools for visualizing the data.
Week 6:
During my sixth week, I compared correlation and diversity. I tried to find a link between the two and my data. The tests were somewhat conclusive, but I need to run more trials and refine the tests before I have truly conclusive evidence. I also rearranged my data sets, and looked at the correlation between sets when the diversity was at its highest.
Week 7:
During my seventh week, I worked on preparing my presentation. I went over the summer's work and consolidated it into slides. I also worked on using a least square estimator to help estimate parameters in my problem. I began programming a script to do so, and researched other statistical estimators in the process.

Presentations


Additional Information