Ragib Morshed email: rmorshed [at] ics.uci.edu

I am a first year Ph.D. student in the Computer Science Department at Donald Bren School of Information and Computer Sciences at University of California at Irvine. My advisor is Professor Charless Fowlkes.

I completed my undergraduate from Pomona College, Claremont, CA, with a double major in Computer Science and Mathematics. My academic and computer science thesis advisor was Tzu-Yi Chen, and my mathematics thesis advisor was Ami Radunskaya.

My research interests are quite broad. My primary research interest is in machine learning, both theory and application. I am also interested in artificial intelligence, compressive sensing, optimization, and computer vision. Even though my research interests are diverse, I have come to discover that all these areas are linked together in interesting ways.

I am a receipient of the ICS Graduate Fellowship in Computer Science for 2009-2012, which is currently supporting my graduate studies.

I recently attended the 2nd Annual Southern California Computer Vision meetup on Friday, October 30, 2009.

Research | Publications | Talks/Presentations | CV | Classes | Contact Info

Research

Machine Learning and Computer Vision

• Currently, I am working on computer vision using machine learning approaches with Professor Charless Fowlkes.
  • Multilinear classifier
    • Multiview face or object detector based on multilinear classifier.
  • Manifold Learning and classification
    • Given a single example of an object, we want to recognize that object from other instances. This is in essence a learning problem from a single example. We are interested in addressing this issue by combining manifold learning with ideas from compressive sensing.
  • Segmentation
    • We are interested in ways of combining bottom up processing like segmentation with high level information about objects, such as recognition of familiar objects or object classes.
• I am part of the computational vision group at the Center for Machine Learning and Intelligent Systems at University of California at Irvine.

Compressive Sensing

• I am still interested in compressive sensing applications to solving problems in computer science, especially in areas of artificial intelligence and machine learning.
• Compressive sensing is an emerging area in signal processing that has become an active area of research in recent years. It basically gives an alternate perspective to the classic system, Ax = b. I got interested in compressive sensing in college after attending a talk given by Emmanuel Candes. I have done some research in this area for my senior theses, and still find it very intriguing.
  
  
  • Undergraduate Senior Thesis in Computer Science
    Department of Computer Science, Pomona College, Claremont, CA.
    Thesis title: Face Recognition in the Real World: A Compressive Sensing Perspective. [pdf]
    Advisor: Tzu-Yi Chen.
    
    • Face recognition is an intriguing area of research in artificial intelligence and computer vision. There are interesting practical applications of face recognition in automatically recognizing and tagging faces in images uploaded to web albums such as Flickr and Picasa. Eigenface is a typcial face recognition technique. We came across a compressive sensing based face recognition technique that was recently developed (2008), and that was shown to be more robust to noise. We ask a number of questions as to the effectiveness of this technique in real world situations, and develop a number of simulations to answer these questions. Some of our answers are surprising, and we provide insights as to the particular answers we obtain.
      
      
  • Undergraduate Senior Thesis in Mathematics
    Department of Mathematics, Pomona College, Claremont, CA.
    Thesis title: Robust Signal Recovery: Designing Stable Measurement Matrices and Random Sensing. [pdf]
    Advisor: Ami Radunskaya.
    
    • An interesting problem in compressive sensing is the design of measurement matrices. Signal recovery is also inadvertently tied to measurement, so measurement matrices are an important consideration in compressive sensing. We explore properties that these matrices must posssess so that they can be effective as measurement matrices in compressive sensing, prove NP-Completness of one of these important properties, and identify various random matrix ensembles that can potentially be used as good measurement matrices. The mathematics behind designing measurement matrices is also closely tied to random matrix theory.
      

Grid Computing

• I spent a summer at The University of Chicago working with the Brain Research Imaging Center and Computation Institute. I was working with Steven Small, Mihael Hategan, and Uri Hasson on a system for management of large-scale science and engineering workflows on grids. I worked on optimizing the grid scheduling system for a workflow system called Swift. I developed a prototype optimization that takes advantage of existing data in grid nodes to probabilistically enhance affinity for such nodes, under certain constraints, for incoming dependent data resulting in lower overall time consumption for workflow processing. It also scales with the size of the grid network.

  A draft of the paper can be found here.

Algorithmic graph theory

• I worked with Professor Tzu-Yi Chen on graph theoretic approaches to non-symmetric perfect elimination matrices. Perfect elimination matrices are important in sparse matrix computations, which show up in solving large systems of the form Ax = b. Perfect elimination has also important consequences in efficient data storage while performing these computations involving sparse matrices.

  We developed a number of small theorems regarding structural properties of graphs that leads to perfect elimination.

Neural Networks

• I am a great fan of neural networks, a powerful and robust technique in machine learning.
  • Choosing and Optimizing Matrix Preconditionars (CHOMP). I worked with America Holloway on her project involving a neural network (written in Perl) to estimate parameters for optimizing the selection of matrix preconditionars.

Physics

• We developed a data logging system in C for accurate measurement of time periods, and studied aerodynamics of various object shapes.

Other Projects

• Here are some other interesting projects that I have worked on.
  
  
  • Mathematical Modeling: Ragib Morshed, Joshua Lippai, and Michael Baumstein. Analysis of the effects of rising sea water levels on Florida's aquifers using an innovative seepage model based on the principles of diusion. Mathematical Contest in Modeling (MCM) 2008 paper. We got a honorable mention though we could not finish all the simulations because of time constraint. [pdf]
    

    Developed a mathematical model in order to determine the effectiveness of terraforming Mars, and estimate values of relevant parameters for feasibility. The model used was based on a simple climate model.

  • Robotics: Developed an autonomous vehicle that uses visual servoing, and sonar, to navigate itself inside a building. Vehicle can also navigate outside with the help of a human guide. Coded a robotic arm movement simulation (using an RRT based algorithm) that was noted as one of the best simulations by Professor Zachary Dodds.
    
    
  • Neural Networks: Developed a game playing agent that plays the board game Mancala. The agent was a multi-layer perceptron, using back propagation learning with temporal difference. Agent was trained using self play. [pdf]
    
    
  • Optimization: Developed an Integer Linear Programming (ILP) approach to innovate, and address the recurring issues with the room draw process for students at Pomona College, Claremont, CA.

Publications

• Yavor Kostov, Ragib Morshed, Barbara Hoeling, Ray Chen, and P. B. Siegel. Period-speed analysis of a pendulum. American Journal of Physics. Vol. 76. No. 10. pg. 956-962. October 2008. [pdf]

Talks/Presentations

• Senior Thesis Presentation, Computer Science Colloquium, Department of Computer Science, Pomona College, 2009. [pdf]
• Senior Thesis Presentation, Department of Mathematics, Pomona College, 2009. [pdf]
• Pomona College Summer Research Poster Conference, Pomona College, 2006, 2007, 2008.
• Physics Colloquium, Department of Physics, Pomona College, 2006.
• Computer Science Colloquium, Department of Computer Science, Pomona College, 2007, 2008.
• Summer Research Poster Presentation, REBMI conference, Claremont, CA, 2008.

Classes

Fall 2009

My classes

• CS 216: Image Understanding.
• CS 276: Belief Networks.
• CS 279S: Seminar in Artificial Intelligence.
• ICS 200: Seminar in Research in ICS.
• ICS 399: University Teaching.

• ICS 6D: Discrete Mathematics for Computer Science.

• Convex Optimization

Contact Info