A fast dual algorithm for kernel logistic regression

This paper gives a new iterative algorithm for kernel logistic regression. It is based on the solution of a dual problem using ideas similar to those of the Sequential Minimal Optimization algorithm for Support Vector Machines. Asymptotic convergence of the algorithm is proved. Computational experiments show that the algorithm is robust and fast. The algorithmic ideas can also be used to give a fast dual algorithm for solving the optimization problem arising in the inner loop of Gaussian Process classifiers.

Large-Scale Elastic Net Regularized Linear Classification SVMs and Logistic Regression

Elastic Net Regularizers have shown much promise in designing sparse classifiers for linear classification. In this work, we propose an alternating optimization approach to solve the dual problems of elastic net regularized linear classification Support Vector Machines (SVMs) and logistic regression (LR). One of the sub-problems turns out to be a simple projection. The other sub-problem can be solved using dual coordinate descent methods developed for non-sparse L2-regularized linear SVMs and LR, without altering their iteration complexity and convergence properties. Experiments on very large datasets indicate that the proposed dual coordinate descent - projection (DCD-P) methods are fast and achieve comparable generalization performance after the first pass through the data, with extremely sparse models.

EEG seizure in bilateral ECT is different between low and high stimulus doses

Melancholic depressive patients referred for ECT were randomized to receive either low dose (n = 20) or high dose (n = 20) stimulus applied bifrontotemporally. The two stimulus groups were comparable on the clinical variables. The EEG seizure was recorded on two channels (right and left frontal), digitized, coded and analyzed offline without knowledge of ECT parameters. EEG seizure was of comparable duration in the two stimulus (high dose and low dose) groups. A new composite measure, Strength-Symmetry-Index (SSI), based on strength and symmetry of seizure EEG was computed using fractal geometry. The SSI of the early-seizure was higher in the high dose than in the low dose ECT group. In a stepwise, logistic regression model, this variable contributed to 65% with correct classification of high dose and low dose ECT seizures.

Climate change induced risk in water quality control problems

A modeling framework is presented in this paper, integrating hydrologic scenarios projected from a General Circulation Model (GCM) with a water quality simulation model to quantify the future expected risk. Statistical downscaling with a Canonical Correlation Analysis (CCA) is carried out to develop the future scenarios of hydro-climate variables starting with simulations provided by a GCM. A Multiple Logistic Regression (MLR) is used to quantify the risk of Low Water Quality (LWQ) corresponding to a threshold quality level, by considering the streamflow and water temperature as explanatory variables. An Imprecise Fuzzy Waste Load Allocation Model (IFWLAM) presented in an earlier study is then used to develop adaptive policies to address the projected water quality risks. Application of the proposed methodology is demonstrated with the case study of Tunga-Bhadra river in India. The results showed that the projected changes in the hydro-climate variables tend to diminish DO levels, thus increasing the future risk levels of LWQ. (C) 2012 Elsevier B.V. All rights reserved.

External support and innovation performance of SMEs in Bangalore: Role of firm level factors

This paper probes the role of internal factors in SMEs in obtaining external support and achieving innovation performance in the context of auto component, electronics and machine tool industries of Bangalore in India. Using step-wise logistic regression analysis, the study found that only if SMEs have internal technical competence in terms of technically qualified entrepreneur, an exclusive design centre, and innovate more frequently, they will be able to obtain external support. Further using step-wise multiple regression the study concluded that SMEs which have come up to implement innovative ideas or exploit market opportunities and which have obtained external support with technically qualified entrepreneurs are able to exhibit better innovation performance.

Characterising weather patterns associated with fire in a seasonally dry tropical forest in southern India

Anthropogenic fires in seasonally dry tropical forests are a regular occurrence during the dry season. Forest managers in India, who presently follow a fire suppression policy in such forests, would benefit from a system of assessing the potential risk to fire on a particular day. We examined the relationship between weather variables (seasonal rainfall, relative humidity, temperature) and days of fire during the dry seasons of 2004-2010, based on MODIS fire incident data in the seasonally dry tropical forests of Mudumalai in the Western Ghats, southern India. Logistic regression analysis showed that high probabilities of a fire day, indicating successful ignition of litter and grass fuel on the forest floor, were associated with low levels of early dry season rainfall, low daily average relative humidity and high daily average temperatures. These weather conditions are representative of low moisture levels of fine fuels, suggesting that the occurrence of fire is moderated by environmental conditions that reduce the flammability of fine fuels in the dry tropics. We propose a quantitative framework for assessing risk of a fire day to assist forest managers in anticipating fire occurrences in this seasonally dry tropical forest, and possibly for those across South Asia.

Surrogate Regret Bounds for Bipartite Ranking via Strongly Proper Losses

The problem of bipartite ranking, where instances are labeled positive or negative and the goal is to learn a scoring function that minimizes the probability of mis-ranking a pair of positive and negative instances (or equivalently, that maximizes the area under the ROC curve), has been widely studied in recent years. A dominant theoretical and algorithmic framework for the problem has been to reduce bipartite ranking to pairwise classification; in particular, it is well known that the bipartite ranking regret can be formulated as a pairwise classification regret, which in turn can be upper bounded using usual regret bounds for classification problems. Recently, Kotlowski et al. (2011) showed regret bounds for bipartite ranking in terms of the regret associated with balanced versions of the standard (non-pairwise) logistic and exponential losses. In this paper, we show that such (non-pairwise) surrogate regret bounds for bipartite ranking can be obtained in terms of a broad class of proper (composite) losses that we term as strongly proper. Our proof technique is much simpler than that of Kotlowski et al. (2011), and relies on properties of proper (composite) losses as elucidated recently by Reid and Williamson (2010, 2011) and others. Our result yields explicit surrogate bounds (with no hidden balancing terms) in terms of a variety of strongly proper losses, including for example logistic, exponential, squared and squared hinge losses as special cases. An important consequence is that standard algorithms minimizing a (non-pairwise) strongly proper loss, such as logistic regression and boosting algorithms (assuming a universal function class and appropriate regularization), are in fact consistent for bipartite ranking; moreover, our results allow us to quantify the bipartite ranking regret in terms of the corresponding surrogate regret. We also obtain tighter surrogate bounds under certain low-noise conditions via a recent result of Clemencon and Robbiano (2011).