My decision to pursue a PhD is derived from my passion for science and engineering paired with my abilities in the field of machine learning and applied statistics. I consider myself fortunate to be part of the Department of Computer Science, University of Florida for my master studies. More importantly, I am glad to have two excellent professors in this field as advisors, Dr. Pader and Dr. Jilson, who are guiding me throughout my graduate studies. They assisted me to decide and pursue the courses and topics that interested me. During my first semester, I took the course Mathematical methods for Intelligent Systems that gave me a strong base for applied mathematics in the field of intelligent systems. Similarly, the research course Computational Neuroscience gave me an insight into applications of statistics, neural networks, and linear dynamical systems in a biological perspective.
My keen interest towards the field of applied statistics, inspired me to take courses such as Machine learning and Neural Networks in the subsequent semester. In this context, I would like to give a brief outline of my master’s research projects, which are I found to be very exciting. The first project was to design a Handwritten Recognition system capable of classifying the digits using the Multilayer Perceptron architecture. Another project was a comparative study of machine learning methodologies such as Bayesian Linear Regression (BLR), Support Vector Machines (SVMs), and Relevance Vector Machines (RVMs), using handwritten character data from postal system. In the first phase, we analyzed the capability of mapping the features calculated on the input character images to membership values in different classes using BLR. In the second phase, the classification capability and sparsity of SVM and RVM are studied.
In addition, I completed an independent study with Dr. Wilson in which I analyzed machine-learning techniques that could be used in a realm-based Question Answering system. My master’s thesis is an extension to this study, for which, I am working with “Morpheus” team at Database Research Center in our department. Our team is building a question answering system that uses deep web sources by exploiting information from Wikipedia and WordNet along with sample query answering strategies provided by users. The motivation behind this research work is as follows.
If we search for an answer to a question in a typical search engines such as Google, Bing, or Yahoo, it usually gives us relevant pages based on the key words in the query. We may need to follow several links or pages to reach a document providing a relevant answer. If we can store such search pathways to an answer for a given user query and reuse it for future searches it may speed up this process. Our question answering system motivated by reuse of prior web search pathways to yield an answer a user query. We represent queries and search pathways in a semi-structured format that contains query...