Variable Sparsity Kernel Learning

This paper(1) presents novel algorithms and applications for a particular class of mixed-norm regularization based Multiple Kernel Learning (MKL) formulations. The formulations assume that the given kernels are grouped and employ l(1) norm regularization for promoting sparsity within RKHS norms of each group and l(s), s >= 2 norm regularization for promoting non-sparse combinations across groups. Various sparsity levels in combining the kernels can be achieved by varying the grouping of kernels-hence we name the formulations as Variable Sparsity Kernel Learning (VSKL) formulations. While previous attempts have a non-convex formulation, here we present a convex formulation which admits efficient Mirror-Descent (MD) based solving techniques. The proposed MD based algorithm optimizes over product of simplices and has a computational complexity of O (m(2)n(tot) log n(max)/epsilon(2)) where m is no. training data points, n(max), n(tot) are the maximum no. kernels in any group, total no. kernels respectively and epsilon is the error in approximating the objective. A detailed proof of convergence of the algorithm is also presented. Experimental results show that the VSKL formulations are well-suited for multi-modal learning tasks like object categorization. Results also show that the MD based algorithm outperforms state-of-the-art MKL solvers in terms of computational efficiency.

Pipelined ring algorithm for matrix multiplication on transputer networks: performance analysis and estimation

A parallel matrix multiplication algorithm is presented, and studies of its performance and estimation are discussed. The algorithm is implemented on a network of transputers connected in a ring topology. An efficient scheme for partitioning the input matrices is introduced which enables overlapping computation with communication. This makes the algorithm achieve near-ideal speed-up for reasonably large matrices. Analytical expressions for the execution time of the algorithm have been derived by analysing its computation and communication characteristics. These expressions are validated by comparing the theoretical results of the performance with the experimental values obtained on a four-transputer network for both square and irregular matrices. The analytical model is also used to estimate the performance of the algorithm for a varying number of transputers and varying problem sizes. Although the algorithm is implemented on transputers, the methodology and the partitioning scheme presented in this paper are quite general and can be implemented on other processors which have the capability of overlapping computation with communication. The equations for performance prediction can also be extended to other multiprocessor systems.

DockYard-a repository to assist modeling of protein-protein docking

In the absence of interlogs, building docking models is a time intensive task, involving generation of a large pool of docking decoys followed by refinement and screening to identify near native docking solutions. This limits the researcher interested in building docking methods with the choice of benchmarking only a limited number of protein complexes. We have created a repository called dockYard (http://pallab.serc.iisc.ernet.in/dockYard), that allows modelers interested in protein-protein interaction to access large volume of information on protein dimers and their interlogs, and also download decoys for their work if they are interested in building modeling methods. dockYard currently offers four categories of docking decoys derived from: Bound (native dimer co-crystallized), Unbound (individual subunits are crystallized, as well as the target dimer), Variants (match the previous two categories in at least one subunit with 100% sequence identity), and Interlogs (match the previous categories in at least one subunit with >= 90% or >= 50% sequence identity). The web service offers options for full or selective download based on search parameters. Our portal also serves as a repository to modelers who may want to share their decoy sets with the community.

Non-viral ex vivo hepatic gene transfer by in situ lipofection of liver and intraperitoneal transplantation of hepatocytes

Perfusion of liver with plasmid DNA-lipofectin complexes via the portal vein results in efficient accumulation of the vector in hepatocytes. Such hepatocytes, when administered intraperitoneally into a hepatectomized rat, repopulate the liver and express the transgene efficiently. This procedure obviates the need for large-scale hepatocyte culture for ex vivo gene transfer. Further, intraperitoneal transplantation is a simple and cost-effective strategy of introducing genetically modified hepatocytes into liver. Thus, in situ lipofection of liver and intraperitoneal transfer of hepatocytes can be developed into a novel method of non-viral ex vivo gene transfer technique that has applications in the treatment of metabolic disorders of liver and hepatic gene therapy.

Accelerating multi-core simulators

Simulation is an important means of evaluating new microarchitectures. With the invention of multi-core (CMP) platforms, simulators are becoming larger and more complex. However, with the availability of CMPs with larger caches and higher operating frequency, the wall clock time required for simulating an application has become comparatively shorter. Reducing this simulation time further is a great challenge, especially in the case of multi-threaded workload due to indeterminacy introduced due to simultaneously executing various threads. In this paper, we propose a technique for speeding multi-core simulation. The model of the processor core and cache are replaced with functional models, to achieve speedup. A timed Petri net model is used to estimate the execution time of the processor and the memory access latencies are estimated using hit/miss information obtained from the functional model of the cache. This model can be used to predict performance of data parallel applications or multiprogramming workload on CMP platform with various cache hierarchies and shared bus interconnect. The error in estimation of the execution time of an application is within 6%. The speedup achieved ranges between an average of 2x--4x over the cycle accurate simulator.

Online character recognition using regression techniques

This paper introduces a scheme for classification of online handwritten characters based on polynomial regression of the sampled points of the sub-strokes in a character. The segmentation is done based on the velocity profile of the written character and this requires a smoothening of the velocity profile. We propose a novel scheme for smoothening the velocity profile curve and identification of the critical points to segment the character. We also porpose another method for segmentation based on the human eye perception. We then extract two sets of features for recognition of handwritten characters. Each sub-stroke is a simple curve, a part of the character, and is represented by the distance measure of each point from the first point. This forms the first set of feature vector for each character. The second feature vector are the coeficients obtained from the B-splines fitted to the control knots obtained from the segmentation algorithm. The feature vector is fed to the SVM classifier and it indicates an efficiency of 68% using the polynomial regression technique and 74% using the spline fitting method.

Power Efficient Redundant Execution for Chip Multiprocessor

This paper describes the design of a power efficient microarchitecture for transient fault detection in chip multiprocessors (CMPs) We introduce a new per-core dynamic voltage and frequency scaling (DVFS) algorithm for our architecture that significantly reduces power dissipation for redundant execution with a minimal performance overhead. Using cycle accurate simulation combined with a simple first order power model, we estimate that our architecture reduces dynamic power dissipation in the redundant core by an mean value of 79% and a maximum of 85% with an associated mean performance overhead of only 1:2%

Extension of TSVM to multi-class and hierarchical text classification problems With general losses

Transductive SVM (TSVM) is a well known semi-supervised large margin learning method for binary text classification. In this paper we extend this method to multi-class and hierarchical classification problems. We point out that the determination of labels of unlabeled examples with fixed classifier weights is a linear programming problem. We devise an efficient technique for solving it. The method is applicable to general loss functions. We demonstrate the value of the new method using large margin loss on a number of multi-class and hierarchical classification datasets. For maxent loss we show empirically that our method is better than expectation regularization/constraint and posterior regularization methods, and competitive with the version of entropy regularization method which uses label constraints.

Contrasting geochemical signatures on land from the Middle and Late Permian extinction events

The end of the Palaeozoic is marked by two mass-extinction events during the Middle Permian (Capitanian) and the Late Permian (Changhsingian). Given similarities between the two events in geochemical signatures, such as large magnitude negative C-13 anomalies, sedimentological signatures such as claystone breccias, and the approximate contemporaneous emplacement of large igneous provinces, many authors have sought a common causal mechanism. Here, a new high-resolution continental record of the Capitanian event from Portal Mountain, Antarctica, is compared with previously published Changhsingian records of geochemical signatures of weathering intensity and palaeoclimatic change. Geochemical means of discriminating sedimentary provenance (Ti/Al, U/Th and La/Ce ratios) all indicate a common provenance for the Portal Mountain sediments and associated palaeosols, so changes spanning the Capitanian extinction represent changes in weathering intensity rather than sediment source. Proxies for weathering intensity chemical index of alteration, W and rare earth element accumulation all decline across the Capitanian extinction event at Portal Mountain, which is in contrast to the increased weathering recorded globally at the Late Permian extinction. Furthermore, palaeoclimatic proxies are consistent with unchanging or cooler climatic conditions throughout the Capitanian event, which contrasts with Changhsingian records that all indicate a significant syn-extinction and post-extinction series of greenhouse warming events. Although both the Capitanian and Changhsingian event records indicate significant redox shifts, palaeosol geochemistry of the Changhsingian event indicates more reducing conditions, whereas the new Capitanian record of reduced trace metal abundances (Cr, Cu, Ni and Ce) indicates more oxidizing conditions. Taken together, the differences in weathering intensity, redox and the lack of evidence for significant climatic change in the new record suggest that the Capitanian mass extinction was not triggered by dyke injection of coal-beds, as in the Changhsingian extinction, and may instead have been triggered directly by the Emeishan large igneous province or by the interaction of Emeishan basalts with platform carbonates.

Whole exome sequencing in an Indian family links Coats plus syndrome and dextrocardia with a homozygous novel CTC1 and a rare HES7 variation

Background: Coats plus syndrome is an autosomal recessive, pleiotropic, multisystem disorder characterized by retinal telangiectasia and exudates, intracranial calcification with leukoencephalopathy and brain cysts, osteopenia with predisposition to fractures, bone marrow suppression, gastrointestinal bleeding and portal hypertension. It is caused by compound heterozygous mutations in the CTC1 gene. Case presentation: We encountered a case of an eight-year old boy from an Indian family with manifestations of Coats plus syndrome along with an unusual occurrence of dextrocardia and situs inversus. Targeted resequencing of the CTC1 gene as well as whole exome sequencing (WES) were conducted in this family to identify the causal variations. The identified candidate variations were screened in ethnicity matched healthy controls. The effect of CTC1 variation on telomere length was assessed using Southern blot. A novel homozygous missense mutation c.1451A > C (p.H484P) in exon 9 of the CTC1 gene and a rare 3'UTR known dbSNP variation (c.*556 T > C) in HES7 were identified as the plausible candidates associated with this complex phenotype of Coats plus and dextrocardia. This CTC1 variation was absent in the controls and we also observed a reduced telomere length in the affected individual's DNA, suggesting its likely pathogenic nature. The reported p.H484P mutation is located in the N-terminal 700 amino acid regionthat is important for the binding of CTC1 to ssDNA through its two OB domains. WES data also showed a rare homozygous missense variation in the TEK gene in the affected individual. Both HES7 and TEK are targets of the Notch signaling pathway. Conclusions: This is the first report of a genetically confirmed case of Coats plus syndrome from India. By means of WES, the genetic variations in this family with unique and rare complex phenotype could be traced effectively. We speculate the important role of Notch signaling in this complex phenotypic presentation of Coats plus syndrome and dextrocardia. The present finding will be useful for genetic diagnosis and carrier detection in the family and for other patients with similar disease manifestations.

Accurate Indoor Positioning of Firefighters using Dual Foot-mounted Inertial Sensors and Inter-agent Ranging

A real-time cooperative localization system, utilizing dual foot-mounted low-cost inertial sensors and RF-based inter-agent ranging, has been developed. Scenario-based tests have been performed, using fully-equipped firefighters mimicking a search operation in a partly smoke-filled environment, to evaluate the performance of the TOR (Tactical lOcatoR) system. The performed tests included realistic firefighter movements and inter-agent distances, factors that are crucial in order to provide realistic evaluations of the expected performance in real-world operations. The tests indicate that the TOR system may be able to provide a position accuracy of approximately two to three meters during realistic firefighter operations, with only two smoke diving firefighters and one supervising firefighter within range.