|
Sinjini Mitra
USC Information Sciences Institute 4676 Admiralty Way, Suite 1001 Marina del Rey, CA 90292.
Phone:     (310) 448-8447
|
CV (pdf format).
| Educational Background |
| Ph.D. |         Carnegie Mellon University (Statistics, August 2005) | |
| M.Stat. |         Indian Statistical Institute, Kolkata, India (Major: Statistics with specialization in Applied Statistics and Data Analysis) | |
| B.Sc. (Hons) |         Presidency College, Kolkata, India (Major: Statistics) |
| Research Interests |
| Current Research Projects |
Studying Learning Capabilities of Students for Building
Intelligent Tutoring Systems
Intelligent tutoring systems have been shown to improve learner
achievement in K-12 educational institutions but students have often
been found to not use these systems very efficiently. For example,
learners may choose random answers (guess), repeatedly request help
until the correct answer is revealed (help abuse) or skip
problems. Moreover, there is growing recognition that student
motivation and engagement must also be considered in addition to
cognitive processes. The present research focuses on classification
of learner action patterns into finite state machines and study how
they differ across different groups of students defined by the
integration of students' self-report data and teacher ratings on of
learner motivation and achievement. The study included high school
students from different schools who worked with a mathematics
tutoring system as part of their regular classroom curriculum. We
adopt a statistical modeling approach based on Hidden Markov Models
(HMM) where the engagement level is used as the hidden
variable. Based on these models, we make inference about the
students' actions and also assess the prediction abilities of the
fitted models by determining accuracies of prediction on future
sessions of students from different school areas. More...
Plan Recognition and Agent Tracking
This project aims at developing a rigorous analytical framework for tracking and detecting malicious activities in large networks involving millions of agents. These agents are engaged in individual and collective activities, often involving deceptive behavior. Most of these agenst are benign and a small fraction of them intend harm. The theoretical foundation of our models is provided by variants of Hidden Markov Models (HMM), in particular, by Abstract Hidden Markov Models (AHMM). Such models are based on hierarchical and graphical representations of causal relationships among various phenomena, in a similar structure as in Dynamic Bayesian Networks (DBN). The test-bed for our models is provided by HATS, a virtual world of benign and covert agents carrying out different plans that involve attacking certain landmarks called beacons. Some particular tasks of interest include (1) tracking groups of agents using a multi-agent scenario by exploiting coordination and associations among different agents who may have similar goals, (2) developing a theory of detection to identify deceptive behavior in the tracking algorithm, (3) making inference by efficient management of a large number of hypotheses. More...
Biometrics (Ph.D. dissertation)
The term biometrics denotes unique biological traits (physical and
behavioral) of individuals that can be used for
identification. The technology of identifying an individual based on
his or her biometrics is termed biometric authentication. My
research on biometrics is in collaboration with Carnegie Mellon
University's CyLab and
focuses primarily on facial biometrics. In particular, I am
intrerested in studying the role of statistical models in developing
rigorous systems and in performance evaluation of such
systems. Another topic that I am interested in is exploring the role
of facial asymmetry in devising efficient face
authentication tools, both using a
spatial domain and a frequency domain representation. More...
| Professional Activities |
| Academic Links |
| Personal Links |