Learning polyhedral classifiers using logistic function

In this paper we propose a new algorithm for learning polyhedral classifiers. In contrast to existing methods for learning polyhedral classifier which solve a constrained optimization problem, our method solves an unconstrained optimization problem. Our method is based on a logistic function based model for the posterior probability function. We propose an alternating optimization algorithm, namely, SPLA1 (Single Polyhedral Learning Algorithm1) which maximizes the loglikelihood of the training data to learn the parameters. We also extend our method to make it independent of any user specified parameter (e.g., number of hyperplanes required to form a polyhedral set) in SPLA2. We show the effectiveness of our approach with experiments on various synthetic and real world datasets and compare our approach with a standard decision tree method (OC1) and a constrained optimization based method for learning polyhedral sets.

And yet it shrinks: A novel method for correcting bias in forest tree growth estimates caused by water-induced fluctuations

Accuracy in tree woody growth estimates is important to global carbon budget estimation and climate-change science. Tree growth in permanent sampling plots (PSPs) is commonly estimated by measuring stem diameter changes, but this method is susceptible to bias resulting from water-induced reversible stem shrinkage. In the absence of bias correction, temporal variability in growth is likely to be overestimated and incorrectly attributed to fluctuations in resource availability, especially in forests with high seasonal and inter-annual variability in water. We propose and test a novel approach for estimating and correcting this bias at the community level. In a 50-ha PSP from a seasonally dry tropical forest in southern India, where tape measurements have been taken every four years from 1988 to 2012, for nine trees we estimated bias due to reversible stem shrinkage as the difference between woody growth measured using tree rings and that estimated from tape. We tested if the bias estimated from these trees could be used as a proxy to correct bias in tape-based growth estimates at the PSP scale. We observed significant shrinkage-related bias in the growth estimates of the nine trees in some censuses. This bias was strongly linearly related to tape-based growth estimates at the level of the PSP, and could be used as a proxy. After bias was corrected, the temporal variance in growth rates of the PSP decreased, while the effect of exceptionally dry or wet periods was retained, indicating that at least a part of the temporal variability arose from reversible shrinkage-related bias. We also suggest that the efficacy of the bias correction could be improved by measuring the proxy on trees that belong to different size classes and census timing, but not necessarily to different species. Our approach allows for reanalysis - and possible reinterpretation of temporal trends in tree growth, above ground biomass change, or carbon fluxes in forests, and their relationships with resource availability in the context of climate change. (C) 2014 Elsevier B.V. All rights reserved.

A comparative study of multiple attribute tree and inverted file structures for large bibliographic files

A variety of data structures such as inverted file, multi-lists, quad tree, k-d tree, range tree, polygon tree, quintary tree, multidimensional tries, segment tree, doubly chained tree, the grid file, d-fold tree. super B-tree, Multiple Attribute Tree (MAT), etc. have been studied for multidimensional searching and related problems. Physical data base organization, which is an important application of multidimensional searching, is traditionally and mostly handled by employing inverted file. This study proposes MAT data structure for bibliographic file systems, by illustrating the superiority of MAT data structure over inverted file. Both the methods are compared in terms of preprocessing, storage and query costs. Worst-case complexity analysis of both the methods, for a partial match query, is carried out in two cases: (a) when directory resides in main memory, (b) when directory resides in secondary memory. In both cases, MAT data structure is shown to be more efficient than the inverted file method. Arguments are given to illustrate the superiority of MAT data structure in an average case also. An efficient adaptation of MAT data structure, that exploits the special features of MAT structure and bibliographic files, is proposed for bibliographic file systems. In this adaptation, suitable techniques for fixing and ranking of the attributes for MAT data structure are proposed. Conclusions and proposals for future research are presented.

Bauer Method of MVDR Spectral Factorization for Pitch Modification in the Source Domain

In our earlier work [1], we employed MVDR (minimum variance distortionless response) based spectral estimation instead of modified-linear prediction method [2] in pitch modification. Here, we use the Bauer method of MVDR spectral factorization, leading to a causal inverse filter rather than a noncausal filter setup with MVDR spectral estimation [1]. Further, this is employed to obtain source (or residual) signal from pitch synchronous speech frames. The residual signal is resampled using DCT/IDCT depending on the target pitch scale factor. Finally, forward filters realized from the above factorization are used to get pitch modified speech. The modified speech is evaluated subjectively by 10 listeners and mean opinion scores (MOS) are tabulated. Further, modified bark spectral distortion measure is also computed for objective evaluation of performance. We find that the proposed algorithm performs better compared to time domain pitch synchronous overlap [3] and modified-LP method [2]. A good MOS score is achieved with the proposed algorithm compared to [1] with a causal inverse and forward filter setup.

Relationship between annual rainfall and tree mortality in a tropical dry forest: Results of a 19-year study at Mudumalai, southern India

Variability in rainfall is known to be a major influence on the dynamics of tropical forests, especially rates and patterns of tree mortality. In tropical dry forests a number of contributing factors to tree mortality, including dry season fire and herbivory by large herbivorous mammals, could be related to rainfall patterns, while loss of water potential in trees during the dry season or a wet season drought could also result in enhanced rates of death. While tree mortality as influenced by severe drought has been examined in tropical wet forests there is insufficient understanding of this process in tropical dry forests. We examined these causal factors in relation to inter-annual differences in rainfall in causing tree mortality within a 50-ha Forest Dynamics Plot located in the tropical dry deciduous forests of Mudumalai, southern India, that has been monitored annually since 1988. Over a 19-year period (1988-2007) mean annual mortality rate of all stems >1 cm dbh was 6.9 +/- 4.6% (range = 1.5-17.5%); mortality rates broadly declined from the smaller to the larger size classes with the rates in stems >30 cm dbh being among the lowest recorded in tropical forest globally. Fire was the main agent of mortality in stems 1-5 cm dbh, elephant-herbivory in stems 5-10 cm dbh, and other natural causes in stems > 10 cm dbh. Elephant-related mortality did not show any relationship to rainfall. On the other hand, fire-related mortality was significantly negatively correlated to quantity of rainfall during the preceding year. Mortality due to other causes in the larger stem sizes was significantly negatively correlated to rainfall with a 2-3-year lag, suggesting that water deficit from mild or prolonged drought enhanced the risk of death but only with a time lag that was greater than similar lags in tree mortality observed in other forest types. In this respect, tropical dry forests growing in regions of high rainfall variability may have evolved greater resistance to rainfall deficit as compared to tropical moist or temperate forests but are still vulnerable to drought-related mortality.

An efficient reduction algorithm for computation of interconnect delay variability for statistical timing analysis in clock tree planning

In this paper, we propose a novel and efficient algorithm for modelling sub-65 nm clock interconnect-networks in the presence of process variation. We develop a method for delay analysis of interconnects considering the impact of Gaussian metal process variations. The resistance and capacitance of a distributed RC line are expressed as correlated Gaussian random variables which are then used to compute the standard deviation of delay Probability Distribution Function (PDF) at all nodes in the interconnect network. Main objective is to find delay PDF at a cheaper cost. Convergence of this approach is in probability distribution but not in mean of delay. We validate our approach against SPICE based Monte Carlo simulations while the current method entails significantly lower computational cost.

Matching sender and receiver: poikilothermy and frequency tuning in a tree cricket

Animals communicate in non-ideal and noisy conditions. The primary method they use to improve communication efficiency is sender-receiver matching: the receiver's sensory mechanism filters the impinging signal based on the expected signal. In the context of acoustic communication in crickets, such a match is made in the frequency domain. The males broadcast a mate attraction signal, the calling song, in a narrow frequency band centred on the carrier frequency (CF), and the females are most sensitive to sound close to this frequency. In tree crickets, however, the CF changes with temperature. The mechanisms used by female tree crickets to accommodate this change in CF were investigated at the behavioural and biomechanical level. At the behavioural level, female tree crickets were broadly tuned and responded equally to CFs produced within the naturally occurring range of temperatures (18 to 27 degrees C). To allow such a broad response, however, the transduction mechanisms that convert sound into mechanical and then neural signals must also have a broad response. The tympana of the female tree crickets exhibited a frequency response that was even broader than suggested by the behaviour. Their tympana vibrate with equal amplitude to frequencies spanning nearly an order of magnitude. Such a flat frequency response is unusual in biological systems and cannot be modelled as a simple mechanical system. This feature of the tree cricket auditory system not only has interesting implications for mate choice and species isolation but may also prove exciting for bio-mimetic applications such as the design of miniature low frequency microphones.

Geometric Decision Tree

In this paper, we present a new algorithm for learning oblique decision trees. Most of the current decision tree algorithms rely on impurity measures to assess the goodness of hyperplanes at each node while learning a decision tree in top-down fashion. These impurity measures do not properly capture the geometric structures in the data. Motivated by this, our algorithm uses a strategy for assessing the hyperplanes in such a way that the geometric structure in the data is taken into account. At each node of the decision tree, we find the clustering hyperplanes for both the classes and use their angle bisectors as the split rule at that node. We show through empirical studies that this idea leads to small decision trees and better performance. We also present some analysis to show that the angle bisectors of clustering hyperplanes that we use as the split rules at each node are solutions of an interesting optimization problem and hence argue that this is a principled method of learning a decision tree.

A sequential dual method for structural SVMs

In many real world prediction problems the output is a structured object like a sequence or a tree or a graph. Such problems range from natural language processing to compu- tational biology or computer vision and have been tackled using algorithms, referred to as structured output learning algorithms. We consider the problem of structured classifi- cation. In the last few years, large margin classifiers like sup-port vector machines (SVMs) have shown much promise for structured output learning. The related optimization prob -lem is a convex quadratic program (QP) with a large num-ber of constraints, which makes the problem intractable for large data sets. This paper proposes a fast sequential dual method (SDM) for structural SVMs. The method makes re-peated passes over the training set and optimizes the dual variables associated with one example at a time. The use of additional heuristics makes the proposed method more efficient. We present an extensive empirical evaluation of the proposed method on several sequence learning problems.Our experiments on large data sets demonstrate that the proposed method is an order of magnitude faster than state of the art methods like cutting-plane method and stochastic gradient descent method (SGD). Further, SDM reaches steady state generalization performance faster than the SGD method. The proposed SDM is thus a useful alternative for large scale structured output learning.

OTCYMIST: Otsu-Canny minimal spanning tree for born-digital images

Text segmentation and localization algorithms are proposed for the born-digital image dataset. Binarization and edge detection are separately carried out on the three colour planes of the image. Connected components (CC's) obtained from the binarized image are thresholded based on their area and aspect ratio. CC's which contain sufficient edge pixels are retained. A novel approach is presented, where the text components are represented as nodes of a graph. Nodes correspond to the centroids of the individual CC's. Long edges are broken from the minimum spanning tree of the graph. Pair wise height ratio is also used to remove likely non-text components. A new minimum spanning tree is created from the remaining nodes. Horizontal grouping is performed on the CC's to generate bounding boxes of text strings. Overlapping bounding boxes are removed using an overlap area threshold. Non-overlapping and minimally overlapping bounding boxes are used for text segmentation. Vertical splitting is applied to generate bounding boxes at the word level. The proposed method is applied on all the images of the test dataset and values of precision, recall and H-mean are obtained using different approaches.

A multichannel sampling method for 2-D finite-rate-of-innovation signals

We address the problem of sampling and reconstruction of two-dimensional (2-D) finite-rate-of-innovation (FRI) signals. We propose a three-channel sampling method for efficiently solving the problem. We consider the sampling of a stream of 2-D Dirac impulses and a sum of 2-D unit-step functions. We propose a 2-D causal exponential function as the sampling kernel. By causality in 2-D, we mean that the function has its support restricted to the first quadrant. The advantage of using a multichannel sampling method with causal exponential sampling kernel is that standard annihilating filter or root-finding algorithms are not required. Further, the proposed method has inexpensive hardware implementation and is numerically stable as the number of Dirac impulses increases.

Synthesis of self-light-scattering wrinkle structured ZnO photoanode by sol-gel method for dye-sensitized solar cells

A special morphological zinc oxide (ZnO) photoanode for dye-sensitized solar cell was fabricated by simple sol-gel drop casting technique. This film shows a wrinkled structure resembling the roots of banyan tree, which acts as an effective self scattering layer for harvesting more visible light and offers an easy transport path for photo-injected electrons. These ZnO electrode of low thickness (similar to 5 mu m) gained an enhanced short-circuit current density of 6.15 mA/cm(2), open-circuit voltage of 0.67 V, fill factor of 0.47 and overall conversion efficiency of 1.97 % under 1 sun illumination. This shows a high conversion efficiency and a superior performance than that of ZnO nanoparticle-based photoanode (eta similar to 1.13 %) of high thickness (similar to 8 mu m).

The variation of N-gamma with footing roughness using the method of characteristics

By using the method of characteristics, the effect of footing-soil interface friction angle (delta) on the bearing capacity factor N-gamma was computed for a strip footing. The analysis was performed by employing a curved trapped wedge under the footing base; this wedge joins the footing base at a distance B-t from the footing edge. For a given footing width (B), the value of B-t increases continuously with a decrease in delta. For delta = 0, no trapped wedge exists below the footing base, that is, B-t/B = 0.5. On the contrary, with delta = phi, the point of emergence of the trapped wedge approaches toward the footing edge with an increase in phi. The magnitude of N-gamma increases substantially with an increase in delta/phi. The maximum depth of the plastic zone becomes higher for greater values of delta/phi. The results from the present analysis were found to compare well with those reported in the literature.

A new two point method of obtaining Cv from a consolidation test: Discussion

Careful study of various aspects presented in the note reveals basic fallacies in the concept and final conclusions.The Authors claim to have presented a new method of determining C-v. However, the note does not contain a new method. In fact, the method proposed is an attempt to generate settlement vs. time data using only two values of (t,8). The Authors have used a rectangular hyperbola method to determine C-v from the predicated 8- t data. In this context, the title of the paper itself is misleading and questionable. The Authors have compared C-v values predicated with measured values, both of them being the results of the rectangular hyperbola method.

A New Method For The Generation Of 5,6-Quinolinedione 5-Methide .5. Synthesis Of Polycyclic Heteroaromatics

Reaction of 6-acetoxy-5-bromomethylquinoline (1c) and 2-bromomethyl-4-(2'-pyridyl)phenyl acetate (2b) with tetrachlorocatechol in acetone in the presence of anhydrous potassium carbonate resulted in the formation of diastereomeric products 3c, 3d, 4e and 4f.