Indian Institute of Technology Bombay

About: Indian Institute of Technology Bombay is a education organization based out in Mumbai, India. It is known for research contribution in the topics: Catalysis & Computer science. The organization has 16756 authors who have published 33588 publications receiving 570559 citations.

Cost-based optimization for magic: algebra and implementation

Abstract: Magic sets rewriting is a well-known optimization heuristic for complex decision-support queries. There can be many variants of this rewriting even for a single query, which differ greatly in execution performance. We propose cost-based techniques for selecting an efficient variant from the many choices.Our first contribution is a practical scheme that models magic sets rewriting as a special join method that can be added to any cost-based query optimizer. We derive cost formulas that allow an optimizer to choose the best variant of the rewriting and to decide whether it is beneficial. The order of complexity of the optimization process is preserved by limiting the search space in a reasonable manner. We have implemented this technique in IBM's DB2 C/S V2 database system. Our performance measurements demonstrate that the cost-based magic optimization technique performs well, and that without it, several poor decisions could be made.Our second contribution is a formal algebraic model of magic sets rewriting, based on an extension of the multiset relational algebra, which cleanly defines the search space and can be used in a rule-based optimizer. We introduce the multiset θ-semijoin operator, and derive equivalence rules involving this operator. We demonstrate that magic sets rewriting for non-recursive SQL queries can be modeled as a sequential composition of these equivalence rules.

Nontangency-Based Lyapunov Tests for Convergence and Stability in Systems Having a Continuum of Equilibria

Abstract: This paper focuses on the stability analysis of systems having a continuum of equilibria. Two notions that are of particular relevance to such systems are convergence and semistability. Convergence is the property whereby every solution converges to a limit point that may depend on the initial condition. Semistability is the additional requirement that all solutions converge to limit points that are Lyapunov stable. We give new Lyapunov-function-based results for convergence and semistability of nonlinear systems. These results do not make assumptions of sign definiteness on the Lyapunov function. Instead, our results use a novel condition based on nontangency between the vector field and invariant or negatively invariant subsets of the level or sublevel sets of the Lyapunov function or its derivative and represent extensions of previously known stability results involving semidefinite Lyapunov functions. To illustrate our results we deduce convergence and semistability of the kinetics of the Michaelis--Menten chemical reaction and the closed-loop dynamics of a scalar system under a universal adaptive stabilizing feedback controller.

Quantitative Prediction of Cold Rolling Textures in Low-CarbonSteel by Means of the Lamel Model

Abstract: Rolling textures of low-carbon steel predicted by full constraints and relaxed constraints Taylor models, as well by a self-consistent model, are quantitatively compared to experimental results. It appears that none of these models really performs well, the best results being obtained by the Pancake model. Anew model (“Lamel model”) is then proposed as a further development of the Pancake model. It treats a stack of two lamella-shaped grains at a time. The new model is described in detail, after which the results obtained for rolling of low-carbon steel are discussed. The prediction of the overall texture now is quantitatively correct. However, the γ-fibre components are better predicted than the α-fibre ones. Finally it is concluded that further work is necessary, as the same kind of success is not guaranteed for other cases, such as rolling of f.c.c, materials.

Urban flood resilience planning and management and lessons for the future: a case study of Mumbai, India

Abstract: Mumbai city, having an area of 437 km2 with a population of 12 million, came to a complete halt owing to the unprecedented rainfall of 994 mm during the 24 hours starting 08:30 on 26 July 2005. At least 419 people (and 16 000 cattle) were killed as a result of the ensuing flash floods and landslides in Mumbai municipal area, and another 216 as a result of flood-related illnesses. Over 100 000 residential and commercial establishments and 30 000 vehicles were damaged. The current paper describes Mumbai's drainage system, the details of the flooding and the measures being taken by the city to mitigate such floods in the future. The Mumbai experience would be helpful for planning response strategies for other large cities to cope with similar events in the future.

Fast and accurate text classification via multiple linear discriminant projections

Abstract: .Support vector machines (SVMs) have shown superb performance for text classification tasks. They are accurate, robust, and quick to apply to test instances. Their only potential drawback is their training time and memory requirement. For n training instances held in memory, the best-known SVM implementations take time proportional to n a, where a is typically between 1.8 and 2.1. SVMs have been trained on data sets with several thousand instances, but Web directories today contain millions of instances that are valuable for mapping billions of Web pages into Yahoo!-like directories. We present SIMPL, a nearly linear-time classification algorithm that mimics the strengths of SVMs while avoiding the training bottleneck. It uses Fisher's linear discriminant, a classical tool from statistical pattern recognition, to project training instances to a carefully selected low-dimensional subspace before inducing a decision tree on the projected instances. SIMPL uses efficient sequential scans and sorts and is comparable in speed and memory scalability to widely used naive Bayes (NB) classifiers, but it beats NB accuracy decisively. It not only approaches and sometimes exceeds SVM accuracy, but also beats the running time of a popular SVM implementation by orders of magnitude. While describing SIMPL, we make a detailed experimental comparison of SVM-generated discriminants with Fisher's discriminants, and we also report on an analysis of the cache performance of a popular SVM implementation. Our analysis shows that SIMPL has the potential to be the method of choice for practitioners who want the accuracy of SVMs and the simplicity and speed of naive Bayes classifiers.