December 17, 2001 (Monday)
9 :0 0 a m - 1 :3 0 p m

INVITED TALKS

9 :0 0 a m - 1 0 :3 0 p m
Dr. Rajendra K. Bera
IBM Global Services India, Bangalore
Quantum Computing

Abstract:
Quantum computers is one of the fascinating technologies in development today. Although the technology has a long way to go before it matures and can be used to build computers for regular use, there have been some spectacular successes in terms of research results, both in hardware and in algorithm development. This paper will describe some of the recent advances made in quantum computing to a nonspecialist audience.

1 0 :3 0 a m - 1 2 :0 0 noon
Dr. Vivek Sarkar
IBM T.J. Watson Research Center, USA
End-to-End Adaptive Optimization: Towards Autonomic Virtual Machines

Abstract:
Considerable experience has been gained recently with adaptive optimization techniques for a single program executing on a single virtual machine. These techniques support on-line profiling and adaptive decision-making for dynamic compilation with increasing levels of optimization and specialization to the program's dynamic context. However, a key limitation of these techniques is that they are oblivious to other programs and virtual machines that might be sharing the same physical resources as the program that they target. This limitation is becoming a serious issue with the advent of recent distributed computing models such as grid computing and web services that are specifically geared towards controlled resource sharing among multiple applications.

In this talk, we give an overview of current technologies for adaptive optimization in a single virtual machine, and then outline challenges that arise for adaptive optimization in a broader end-to-end context of multiple virtual machines. We conclude by outlining possible directions for addressing these challenges, including the development of autonomic virtual machines that are capable of self-diagnosis, self-optimization, and self-healing.

1 2 :0 0 noon - 1 :3 0 p m
Dr M. S. Santhanam
IBM India Research Laboratory, New Delhi, India
SVD as a Tool to Understand Large Scientific Data : New Results and Computational Challenges

Abstract:
Scientific and economic data, which appear in the form of time-series, like the ECG data, the daily changes in atmospheric variables, the variations in stock prices etc., are voluminous and the immediate question is how to make sense out of them. Often, singular value decomposition and its variants have been widely used as techniques for dimensionality reduction of such voluminous data. Yet, assigning physical significance to reduced dimensions of the data has always been an outstanding question. Recent developments indicate that ideas based on random matrix theory (RMT) might have the answer. It is a classic case interdisciplinary approach, with ideas like RMT, that originated from studies on complex quantum systems, helping us understand phenomena as disparate as the stock market fluctuations and the atmospheric anomolies over the Atlantic ocean. The bulwark behind such new results are the powerful computational techniques that have come up in the recent years for large eigenvalue problems. In this talk, we will describe these new results with illustrations from atmospheric correlations and stock price fluctuations and point out its applications to image processing. The computational challenges that accompany the application of these new techniques and the role of LAPACK and ESSL libraries and their parallel versions in simplifying the eigensolver computations and improvisation that could enhance the computational efficiency will be discussed. We will also take a peep into the future in the context of emerging high performance computational paradigms like the grid computing.