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.

Robust Formulations for Handling Uncertainty in Kernel Matrices

We study the problem of uncertainty in the entries of the Kernel matrix, arising in SVM formulation. Using Chance Constraint Programming and a novel large deviation inequality we derive a formulation which is robust to such noise. The resulting formulation applies when the noise is Gaussian, or has finite support. The formulation in general is non-convex, but in several cases of interest it reduces to a convex program. The problem of uncertainty in kernel matrix is motivated from the real world problem of classifying proteins when the structures are provided with some uncertainty. The formulation derived here naturally incorporates such uncertainty in a principled manner leading to significant improvements over the state of the art. 1.

Greenhouse gas inventory estimates for India

This article reports the greenhouse gas emissions of anthropogenic origin by sources and removals by sinks of India for 2007 prepared under the aegis of the Indian Network for Climate Change Assessment (INCCA) (note 1). The emission profile includes carbon dioxide (CO(2)), methane and nitrous oxide. It also includes the estimates of hydrofluorocarbons, perfluorocarbons and sulphur hexafluoride at the national level from various sectors, viz, energy, industrial process and product use, agriculture, land-use, land-use change and forestry (LULUCF), and waste. In 2007, emissions were of the order of 2008.67 Tg (note 2) of CO(2) equivalents without emissions from the LULUCF sector. Whereas with LULUCF the emissions were about 1831.65 Tg CO(2) equivalents. The energy sector accounted for 69% of the total emissions, the agriculture sector contributed 19% of the emissions, 9% of the emissions was from the industrial processes and product use, and only 3% of the emissions was attributable to the waste sector. The LULUCF sector on the whole was net sink category for CO(2). The study tracks the improvements made in inventory estimates at the national level through the years, in terms of the expanding coverage of sources, reducing uncertainties and inclusion of new methodologies, including some elements of future areas of work.

Mathematical Model of Viral Kinetics In Vitro Estimates the Number of E2-CD81 Complexes Necessary for Hepatitis C Virus Entry

Interaction between the hepatitis C virus (HCV) envelope protein E2 and the host receptor CD81 is essential for HCV entry into target cells. The number of E2-CD81 complexes necessary for HCV entry has remained difficult to estimate experimentally. Using the recently developed cell culture systems that allow persistent HCV infection in vitro, the dependence of HCV entry and kinetics on CD81 expression has been measured. We reasoned that analysis of the latter experiments using a mathematical model of viral kinetics may yield estimates of the number of E2-CD81 complexes necessary for HCV entry. Here, we constructed a mathematical model of HCV viral kinetics in vitro, in which we accounted explicitly for the dependence of HCV entry on CD81 expression. Model predictions of viral kinetics are in quantitative agreement with experimental observations. Specifically, our model predicts triphasic viral kinetics in vitro, where the first phase is characterized by cell proliferation, the second by the infection of susceptible cells and the third by the growth of cells refractory to infection. By fitting model predictions to the above data, we were able to estimate the threshold number of E2-CD81 complexes necessary for HCV entry into human hepatoma-derived cells. We found that depending on the E2-CD81 binding affinity, between 1 and 13 E2-CD81 complexes are necessary for HCV entry. With this estimate, our model captured data from independent experiments that employed different HCV clones and cells with distinct CD81 expression levels, indicating that the estimate is robust. Our study thus quantifies the molecular requirements of HCV entry and suggests guidelines for intervention strategies that target the E2-CD81 interaction. Further, our model presents a framework for quantitative analyses of cell culture studies now extensively employed to investigate HCV infection.

Time Series Classification for Online Tamil Handwritten Character Recognition – A Kernel Based Approach

In this paper, we consider the problem of time series classification. Using piecewise linear interpolation various novel kernels are obtained which can be used with Support vector machines for designing classifiers capable of deciding the class of a given time series. The approach is general and is applicable in many scenarios. We apply the method to the task of Online Tamil handwritten character recognition with promising results.

Ganga-Brahmaputra river discharge from Jason-2 radar altimetry: An update to the long-term satellite-derived estimates of continental freshwater forcing flux into the Bay of Bengal

This paper discusses the use of Jason-2 radar altimeter measurements to estimate the Ganga-Brahmaputra surface freshwater flux into the Bay of Bengal for the period mid-2008 to December 2011. A previous estimate was generated for 1993-2008 using TOPEX-Poseidon, ERS-2 and ENVISAT, and is now extended using Jason-2. To take full advantages of the new availability of in situ rating curves, the processing scheme is adapted and the adjustments of the methodology are discussed here. First, using a large sample of in situ river height measurements, we estimate the standard error of Jason-2-derived water levels over the Ganga and the Brahmaputra to be respectively of 0.28 m and 0.19 m, or less than similar to 4% of the annual peak-to-peak variations of these two rivers. Using the in situ rating curves between water levels and river discharges, we show that Jason-2 accurately infers Ganga and Brahmaputra instantaneous discharges for 2008-2011 with mean errors ranging from similar to 2180 m(3)/s (6.5%) over the Brahmaputra to similar to 1458 m(3)/s (13%) over the Ganga. The combined Ganga-Brahmaputra monthly discharges meet the requirements of acceptable accuracy (15-20%) with a mean error of similar to 16% for 2009-2011 and similar to 17% for 1993-2011. The Ganga-Brahmaputra monthly discharge at the river mouths is then presented, showing a marked interannual variability with a standard deviation of similar to 12500 m(3)/s, much larger than the data set uncertainty. Finally, using in situ sea surface salinity observations, we illustrate the possible impact of extreme continental freshwater discharge event on the northern Bay of Bengal as observed in 2008.

Efficient Methods for Robust Classification Under Uncertainty in Kernel Matrices

In this paper we study the problem of designing SVM classifiers when the kernel matrix, K, is affected by uncertainty. Specifically K is modeled as a positive affine combination of given positive semi definite kernels, with the coefficients ranging in a norm-bounded uncertainty set. We treat the problem using the Robust Optimization methodology. This reduces the uncertain SVM problem into a deterministic conic quadratic problem which can be solved in principle by a polynomial time Interior Point (IP) algorithm. However, for large-scale classification problems, IP methods become intractable and one has to resort to first-order gradient type methods. The strategy we use here is to reformulate the robust counterpart of the uncertain SVM problem as a saddle point problem and employ a special gradient scheme which works directly on the convex-concave saddle function. The algorithm is a simplified version of a general scheme due to Juditski and Nemirovski (2011). It achieves an O(1/T-2) reduction of the initial error after T iterations. A comprehensive empirical study on both synthetic data and real-world protein structure data sets show that the proposed formulations achieve the desired robustness, and the saddle point based algorithm outperforms the IP method significantly.

Revision of the scolopendrid centipede digitipes Attems, 1930, from India (Chilopoda: Scolopendromorpha): reconciling molecular and morphological estimates of species diversity

Recent work on molecular phylogenetics of Scolopendridae from the Western Ghats, Peninsular India, has suggested the presence of six cryptic species of the otostigmine Digitipes Attems, 1930, together with three species described in previous taxonomic work by Jangi and Dass (1984). Digitipes is the correct generic attribution for a monophyletic group of Indian species, these being united with three species from tropical Africa (including the type) that share a distomedial process on the ultimate leg femur of males that is otherwise unknown in Otostigminae. Second maxillary characters previously used in the diagnosis of Digitipes are dismissed because Indian species do not possess the putatively diagnostic character states. Two new species from the Western Ghats that correspond to groupings identified based on monophyly, sequence divergence and coalescent analysis using molecular data are diagnosed based on distinct morphological characters. They are D. jangii and D. periyarensis n. spp. Three species named by Jangi and Dass (Digitipes barnabasi, D. coonoorensis and D. indicus) are revised based on new collections; D. indicus is a junior subjective synonym of Arthrorhabdus jonesii Verhoeff, 1938, the combination becoming Digitipes jonesii (Verhoeff, 1938) n. comb. The presence of Arthrorhabdus in India is accordingly refuted. Three putative species delimited by molecular and ecological data remain cryptic from the perspective of diagnostic morphological characters and are presently retained in D. barnabasi, D. jangii and D. jonesii. A molecularly-delimited species that resolved as sister group to a well-supported clade of Indian Digitipes is identified as Otostigmus ruficeps Pocock, 1890, originally described from a single specimen and revised herein. One Indian species originally assigned to Digitipes, D. gravelyi, deviates from confidently-assigned Digitipes with respect to several characters and is reassigned to Otostigmus, as O. gravelyi (Jangi and Dass, 1984) n. comb.

A h-Adaptive Algorithm Using Residual Error Estimates for Fluid Flows

Algorithms for adaptive mesh refinement using a residual error estimator are proposed for fluid flow problems in a finite volume framework. The residual error estimator, referred to as the R-parameter is used to derive refinement and coarsening criteria for the adaptive algorithms. An adaptive strategy based on the R-parameter is proposed for continuous flows, while a hybrid adaptive algorithm employing a combination of error indicators and the R-parameter is developed for discontinuous flows. Numerical experiments for inviscid and viscous flows on different grid topologies demonstrate the effectiveness of the proposed algorithms on arbitrary polygonal grids.

Accurate performance analysis of single and opportunistic AF relay cooperation with imperfect cascaded channel estimates

Given the significant gains that relay-based cooperation promises, the practical problems of acquisition of channel state information (CSI) and the characterization and optimization of performance with imperfect CSI are receiving increasing attention. We develop novel and accurate expressions for the symbol error probability (SEP) for fixed-gain amplify-and-forward relaying when the destination acquires CSI using the time-efficient cascaded channel estimation (CCE) protocol. The CCE protocol saves time by making the destination directly estimate the product of the source-relay and relay-destination channel gains. For a single relay system, we first develop a novel SEP expression and a tight SEP upper bound. We then similarly analyze an opportunistic multi-relay system, in which both selection and coherent demodulation use imperfect estimates. A distinctive aspect of our approach is the use of as few simplifying approximations as possible, which results in new results that are accurate at signal-to-noise-ratios as low as 1 dB for single and multi-relay systems. Using insights gleaned from an asymptotic analysis, we also present a simple, closed-form, nearly-optimal solution for allocation of energy between pilot and data symbols at the source and relay(s).

Joint Antenna Selection and Frequency-Domain Scheduling in OFDMA Systems with Imperfect Estimates from Dual Pilot Training Scheme

Transmit antenna selection (AS) has been adopted in contemporary wideband wireless standards such as Long Term Evolution (LTE). We analyze a comprehensive new model for AS that captures several key features about its operation in wideband orthogonal frequency division multiple access (OFDMA) systems. These include the use of channel-aware frequency-domain scheduling (FDS) in conjunction with AS, the hardware constraint that a user must transmit using the same antenna over all its assigned subcarriers, and the scheduling constraint that the subcarriers assigned to a user must be contiguous. The model also captures the novel dual pilot training scheme that is used in LTE, in which a coarse system bandwidth-wide sounding reference signal is used to acquire relatively noisy channel state information (CSI) for AS and FDS, and a dense narrow-band demodulation reference signal is used to acquire accurate CSI for data demodulation. We analyze the symbol error probability when AS is done in conjunction with the channel-unaware, but fair, round-robin scheduling and with channel-aware greedy FDS. Our results quantify how effective joint AS-FDS is in dispersive environments, the interactions between the above features, and the ability of the user to lower SRS power with minimal performance degradation.

Estimates of aerosol absorption over India using multi-satellite retrieval

Aerosol absorption is poorly quantified because of the lack of adequate measurements. It has been shown that the Ozone Monitoring Instrument (OMI) aboard EOS-Aura and the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard EOS-Aqua, which fly in formation as part of the A-train, provide an excellent opportunity to improve the accuracy of aerosol retrievals. Here, we follow a multi-satellite approach to estimate the regional distribution of aerosol absorption over continental India for the first time. Annually and regionally averaged aerosol single-scattering albedo over the Indian landmass is estimated as 0.94 +/- 0.03. Our study demonstrates the potential of multi-satellite data analysis to improve the accuracy of retrieval of aerosol absorption over land.

REPRODUCING KERNEL FOR A CLASS OF WEIGHTED BERGMAN SPACES ON THE SYMMETRIZED POLYDISC

A natural class of weighted Bergman spaces on the symmetrized polydisc is isometrically embedded as a subspace in the corresponding weighted Bergman space on the polydisc. We find an orthonormal basis for this subspace. It enables us to compute the kernel function for the weighted Bergman spaces on the symmetrized polydisc using the explicit nature of our embedding. This family of kernel functions includes the Szego and the Bergman kernel on the symmetrized polydisc.