|
 About Us
Mission & Vision
Faculty
Staff
Students
Computing Facilities
News
Seminars
Contact Us
Research
Laboratories
CCT
Academics
Admission
Graduate
Undergraduate
Courses
Rm Reservation
|
|
Transforming Academic Research Into Commercial Applications |
Transforming Academic Research Into Commercial Applications
Google Co-Founder - Larry Page mentioned in a recent American Association for the Advancement of
Science Conference at San Francisco that Science has a marketing problem.
"There may be the best products one creates in
the lab, but without marketing it becomes a failure if no one knows about it. The Scientists should expend part
of their research budget in marketing and spread the word. Scientists need more entrepreneurial drive
and could benefit by doing more to promote solutions to big human problems", Google Inc. co-founder
Larry Page told a meeting of academic researchers.
Sample Projects
|
Project 1: Fast Web Page Allocation On a Server Using Self-Organizing Properties of Neural
Networks |
Project 1:
Title: Fast Web Page Allocation On a Server Using Self-Organizing Properties of Neural Networks
Authors: Vir V. Phoha (La Tech), S. S. Iyengar (LSU), R. Kannan (LSU)
Description
We have developed a fast competitive neural network learning approach to allocate Web page requests to a cluster of
Web servers. Our approach performs much better as compared to current approaches to route requests among the distributed
Web-server nodes: (1) Client-based, (2) DNS-based, (3) Dispatcher based, and (4) Server based. The optimality of our
allocator is obtained through a learning rule where a network adapts to unpredictable changes using the framework of
competitive learning. The stability of responses are generated through a process of learning. Results show an order of
magnitude improvement over traditional DNS based routing and load-balancing approaches.
Advantages
- Very fast direction of Client requests for Web pages to appropriate Web server.
- The system is scalable to any number of users without performance degradation because a separate connection
is maintained for each user and the connection is loaded into memory as the user is authenticated.
- The system learns the request patterns and adapts to changing behavior of Web traffic.
- System is easy-to-use and cost-effective.
- Automatic enhancement or relaxation of security based on the tunable parameters.
Performance Summary:
Our algorithms ranged between 85% to 98% hit rate compared to a performance range of 2% to 40% hit rate for a round
robin scheme when simulating the real Web traffic. As the traffic increases, our algorithm performs much better than
the Round Robin scheme. These results exceed those of existing systems.
Areas of Application
This technology will be of use to businesses that maintain Websites containing multiple data on more than one server
to do online processing of requests. Examples are most medium to large scale Web retailers.
Patent Status
One patent allowed; second patent pending.
Reference
V. V. Phoha, S. S. Iyengar and R. Kannan, Fast Web Page Allocation With Neural Networks, IEEE Internet Computing,
6 (2002), pp. 18-26.
|
|
Project 2: Adaptive Neural Network Clustering of Web Users |
Project 2:
Title: Adaptive Neural Network Clustering of Web Users
Description
A neural network based on adaptive resonance theory dynamically groups users based on their Web access patterns.
A prefetching application of this clustering technique showed prediction accuracy as high as 97.78 percent. The
degree of personalization that a Web site offers in presenting its services to users is an important attribute
contributing to the site's popularity. Web. Many systems succeed in grouping users according to their diverse
Web interests, they lack the ability to adapt to changes in those interests over time.
Areas of Application
This technology will be of use to businesses that target marketing and products to users with specific interests
and browsing patterns. Examples are most medium to large scale Web retailers.
Reference
S. K. Rangarajan, V. V. Phoha, K. Balagani, R. R. Selmic and S. S. Iyengar, Adaptive Neural Network Clustering of
Web Users, IEEE Computer, 37 (2004), pp. 34-40.
|
|
Project 3: Advanced Data Fusion and Decision Making Using Hyperspectral Images for Diabetic
Retinotherapy |
Project 3
Title: Advanced Data Fusion and Decision Making Using Hyperspectral Images for Diabetic Retinotherapy
Authors:
- Dr. S.S. Iyengar, Dr. Peter Wolenski, Dr. Nathan Brenner, Dr. Karki, Dr. Silvermann (LSU).
- Dr. Hilary Thompson, Dr. Khoobehi, Dr. Roger (LSU Eye Center, New Orleans, LA).
- Dr. Madhu Subramaniam, University of California, San Diego.
- Dr. Vir Phoha (LaTech).
- Dr. Reddy (Grambling University).
Description
This project involves the analysis of combined medical images of anatomical details that represent metabolic activities.
The goals of the analysis are to 1) understand the functional anatomy for early detection of disease, 2) gain knowledge
about the stage and condition of disease, and 3) devise a refined and detailed treatment plan in the presence of disease.
As a first example, the Positron Emission Tomography (PET) and Computed Tomography (CT) scans reveal complementary images
that can be combined for cancer radiation treatment and planning. The CT scan captures clear anatomical details of the
organ, but often cannot detect the disease until it has significantly matured. The PET scan, on the other hand, captures
the functional property of the metabolic rate, and thus reveals the presence of disease without identifying its precise
location. A method that uses both scans efficiently will be a powerful diagnostic tool. A second example involves
analyzing images of the retina obtained from the fundus camera in combination with hyperspectral data. Hyperspectral
images identify potential sites of neovascularization, which is a symptom of Diabetic Retinopathy where oxygen concentration
in the retinal tissues decreases causing new vessel growth. It is envisioned that functional information (the oxygen
saturation level) combined with anatomical features of the fundus camera will assist in diagnosing Diabetic Retinopathy
and treatment planning using Panretinal coagulation.
Preliminary work:
Generating a composite image from multiple image modalities such as PET and CT involves spatially aligning the images so
that the functional and anatomical details are spatially relevant and correlated. An image alignment algorithm commonly
referred to as image registration, will account for the scaling, rotation and translation of features in the images.
More Information
|
|
Project 4: An Image Guided Diagnostic Library for Clinical and Research Use |
Project 4
Title: An Image Guided Diagnostic Library for Clinical and Research Use
Authors:
- Dr. S.S. Iyengar, Dr. Peter Wolenski, Dr. Nathan Brenner, Dr. Karki (LSU).
- Dr. Hilary Thompson, Dr. Khoobehi (LSU Eye Center, New Orleans, LA).
- Dr. Madhu Subramaniam, University of California, San Diego.
Description
The focus of this research is the establishment of national spectral profile library handling large volumes
of image data, combining spectral and spatial matching techniques and efficient data storage and search mechanisms.
|
|
Project 5: Confocal-4D: An Architecture for Real-time Tracking and Volume Rendering of White-light Confocal
Microscope Optical Serial Sections |
Project 5
Title: Confocal-4D: An Architecture for Real-time Tracking and Volume Rendering of White-light Confocal
Microscope Optical Serial Sections
Authors: Madhusudhanan Balasubramanian, Juan Reynaud, Roger W. Beuerman, S. Sitharama Iyengar, Bijaya B. Karki.
Description
Application of this invention (Patent is being pursued on this with LSU and LSU Eye Center)
Application of the invention that has direct use in studying glaucomatous progression in the optic nerve head in the
eye; this is not mentioned in the initial claim or scope of the invention. This may facilitate a superior glaucomatous
progression detection and analysis.
|
|
Project 6: Improved query processing using fuzzy BBC codes |
Project 6
Title: Improved query processing using fuzzy BBC codes
Authors: PI: Lt Col Leemon Baird, PhD (US Air Force), Co-PI: Dr. Donald H. Kraft (LSU)
Description
This research will explore the use of a new type of code, BBC codes, for improving query processing. It will
look at both theoretical analysis and empirical tests of the systems involved, using large databases.
The work can be applied to improving traditional crisp queries (e.g. "A and (B or not C)"), extensions of those
queries ("Majority of A, B, C, D, E"), and fuzzy weighted queries ("A (0.9) and B (0.3)"), preranked queries
("A and B with outputs ranked by Google pagerank"), and various combinations of these. These can also be combined
with each other, and with techniques such as automatically generating queries from user-defined lists of relevant
and non-relevant queries. If successful, this work will have applications to Web searches, queries over specialized
databases (e.g. U.S. patents), and queries formed automatically by data mining systems as part of algorithms for detecting
patterns of interest. Empirical results will compare search speed with traditional approaches, and will explore
the effect of various parameter choices (e.g. using partitioned and hierarchical hash functions in the algorithm
rather than simple hash functions).
|
|
Project 7: UniPACS for Medical Imaging |
Project 7
Title: LSU researchers bring new developments in business and medicine to Louisiana
Authors: Dr. John Tyler (LSU) Co-Founder
Description
LSU CAPITAL, the LSU Health Sciences Center and the Department of LSU Computer Science have joined
forces to create Universal Picture Archiving and Communication Systems, or UniPACS, a new company that
can change medical imaging. The group has created a software package, also known as UniPACS that helps
radiologists view and archive medical images like X-rays, MRIs, and CT scans in a much more cost efficient
and practical way.
With this kind of software, a Louisiana physician or radiologist can offer an additional medical opinion
from more than a thousand miles away by observing an MRI scan performed at another facility through his
wireless laptop computer. The first working model in 2004 went into Charity Hospital in New Orleans,
which lost an undetermined amount of paper records during Hurricane Katrina.
Historically, radiologists have relied on expensive and cumbersome equipment to read the digital image
files that helped them to diagnose patients. With LSU's new UniPACS software, doctors can now view medical
images from any PC with an Internet connection. The new software works as a secure, confidential,
FDA-approved workstation, allowing radiologists to analyze images from almost any location.
Dr. John Tyler
UniPACS
|
|
|
