Interval Data Classification under Partial Information: A Chance-Constraint Approach

This paper presents a Chance-constraint Programming approach for constructing maximum-margin classifiers which are robust to interval-valued uncertainty in training examples. The methodology ensures that uncertain examples are classified correctly with high probability by employing chance-constraints. The main contribution of the paper is to pose the resultant optimization problem as a Second Order Cone Program by using large deviation inequalities, due to Bernstein. Apart from support and mean of the uncertain examples these Bernstein based relaxations make no further assumptions on the underlying uncertainty. Classifiers built using the proposed approach are less conservative, yield higher margins and hence are expected to generalize better than existing methods. Experimental results on synthetic and real-world datasets show that the proposed classifiers are better equipped to handle interval-valued uncertainty than state-of-the-art.

Management objectives of Queensland fisheries: Putting the horse before the cart

A review of future management arrangements for the Queensland East Coast Trawl fishery was undertaken in 2010 to develop a management plan for the next 10 years. A key question raised at the start of the review process was: what should the management plan achieve? As with fisheries management in most countries, multiple management objectives were implicit in policy statements, but were poorly specified in some areas (particularly social objectives) and strongly identified in others (e.g., an objective of sustainability). As a start to the management review process, an analysis of what objectives the management system should aim to achieve was undertaken. A review of natural resource management objectives employed internationally was used to develop a candidate list, and the objectives most relevant to the fishery were short-listed by a scientific advisory group. Additional objectives specific to Queensland fisheries management, but not identified in the international review, were also identified and incorporated into the objective set. The relative importance of the different objectives to different stakeholder groups was assessed using the Analytic Hierarchy Process. As with other studies, the relative importance of the different objectives varied both within and between the different stakeholder groups, although general trends in preferences were observed.

IPM in the transgenic era: a review of the challenges from emerging pests in Australian cotton systems

The Cotton Catchment Communities Cooperative Research Centre began during a period of rapid uptake of Bollgard II® cotton, which contains genes to express two Bt proteins that control the primary pests of cotton in Australia, Helicoverpa armigera and H. punctigera. The dramatic uptake of this technology presumably resulted in strong selection pressure for resistance in Helicoverpa spp. against the Bt proteins. The discovery of higher than expected levels of resistance in both species against one of the proteins in Bollgard II® cotton (Cry2Ab) led to significant re-evaluation of the resistance management plan developed for this technology, which was a core area of research for the Cotton CRC. The uptake of Bollgard II® cotton also led to a substantial decline in pesticide applications against Helicoverpa spp. (from 10–14 to 0–3 applications per season). The low spray environment allowed some pests not controlled by the Bt proteins to emerge as more significant pests, especially sucking species such as Creontiades dilutus and Nezara viridula. A range of other minor pests have also sporadically arisen as problems. Lack of knowledge and experience with these pests created uncertainty and encouraged insecticide use, which threatened to undermine the gains made with Bollgard II® cotton. Here we chronicle the achievements of the Cotton CRC in providing the industry with new knowledge and management strategies for these pests.

Impact of nitrification inhibitor (DMPP) on soil nitrous oxide emissions from an intensive broccoli production system in sub-tropical Australia

Vegetable cropping systems are often characterised by high inputs of nitrogen fertiliser. Elevated emissions of nitrous oxide (N2O) can be expected as a consequence. In order to mitigate N2O emissions from fertilised agricultural fields, the use of nitrification inhibitors, in combination with ammonium based fertilisers, has been promoted. However, no data is currently available on the use of nitrification inhibitors in sub-tropical vegetable systems. A field experiment was conducted to investigate the effect of the nitrification inhibitor 3,4-dimethylpyrazole phosphate (DMPP) on N2O emissions and yield from broccoli production in sub-tropical Australia. Soil N2O fluxes were monitored continuously (3 h sampling frequency) with fully automated, pneumatically operated measuring chambers linked to a sampling control system and a gas chromatograph. Cumulative N2O emissions over the 5 month observation period amounted to 298 g-N/ha, 324 g-N/ha, 411 g-N/ha and 463 g-N/ha in the conventional fertiliser (CONV), the DMPP treatment (DMPP), the DMMP treatment with a 10% reduced fertiliser rate (DMPP-red) and the zero fertiliser (0N), respectively. The temporal variation of N2O fluxes showed only low emissions over the broccoli cropping phase, but significantly elevated emissions were observed in all treatments following broccoli residues being incorporated into the soil. Overall 70–90% of the total emissions occurred in this 5 weeks fallow phase. There was a significant inhibition effect of DMPP on N2O emissions and soil mineral N content over the broccoli cropping phase where the application of DMPP reduced N2O emissions by 75% compared to the standard practice. However, there was no statistical difference between the treatments during the fallow phase or when the whole season was considered. This study shows that DMPP has the potential to reduce N2O emissions from intensive vegetable systems, but also highlights the importance of post-harvest emissions from incorporated vegetable residues. N2O mitigation strategies in vegetable systems need to target these post-harvest emissions and a better evaluation of the effect of nitrification inhibitors over the fallow phase is needed.

Aggregation studies in crystals of apolar helical peptides: Boc-Aib-Val-Ala-Leu-Aib-Val-Ala-Leu-Aib-OMe

In the crystal, the backbone of Boc-(Aib-Val-Ala-Leu)2-Aib-OMe adopts a helical form with four alpha-type hydrogen bonds in the middle, flanked by 3(10)-type hydrogen bonds at either end. The helical molecules stack in columns with head-to-tail hydrogen bonds, either directly between NH and CO, or bridged by solvent molecules. The packing of the helices is parallel, even in space group P2(1). Cell parameters are a = 9.837(2) A, b = 15.565(3) A, c = 20.087(5) A, beta = 96.42(2) degrees, dcalc = 1.091 g/cm3 for C46H83N9O12.1.5H2O.0.67CH3OH. There appears to be some hydration of the backbone in this apolar helix.

Modular design of synthetic protein mimics. Crystal structure of two seven-residue helical peptide segments linked by .epsilon.-aminocaproic acid

Two seven-residue helical segments, Val-Ala-Leu-Aib-Val-Ala-Leu, were linked synthetically with an epsilon-aminocaproic acid (Acp) linker with the intention of making a stable antiparallel helix-helix motif. The crystal structure of the linked peptide Boc-Val-Ala-Leu-Aib-Val-Ala-Leu-Acp-Val-Ala-Leu-Aib-Val-Ala-Leu-OMe (1) shows the two helices displaced laterally from each other by the linker, but the linker has not folded the molecule into a close-packed antiparallel conformation. Two strong intermolecular NH...O = C hydrogen bonds are formed between the top of the lower helix of one molecule and the bottom of the upper helix in a laterally adjacent molecule to give the appearance of an extended single helix. The composite peptide with Boc and OMe end groups, C76H137N15O18.H2O, crystallize in space group P2(1) with a = 8.802 (1) angstrom, b = 20.409 (4) angstrom, c = 26.315 (3) angstrom, and beta = 90.72 (1)degrees; overall agreement R = 7.86% for 5030 observed reflections (\F(o)\ > 3-sigma(F)); resolution = 0.93 angstrom. Limited evidence for a more compact conformation in solution consistent with an antiparallel helix arrangement is obtained by comparison of the HPLC retention times and CD spectra of peptide 1 with well-characterized continuous helices of similar length and sequence.

Zervamicins, a structurally characterised peptide model for membrane ion channels

Voltage dependent membrane channels are formed by the zervamicins, a group of α-aminoisobutyric acid containing peptides. The role of polar residues like Thr, Gln and Hyp in promoting helical bundle formation is established by dramatically reduced channel lifetimes for a synthetic apolar analog. Crystal structures of Leu1-zervamicin reveal association of bent helices. Polar contacts between convex faces result in an ‘hour glass’ like arrangement of an aqueous channel with a central constriction. The structure suggests that gating mechanisms may involve movement of the Gln11 carboxamide group. Gln3 may play a role in modulating the size of the channel mouth.

Peptide mimics for structural features in proteins. Crystal structures of three heptapeptide helices with a C-terminal 6-->1 hydrogen bond.

The crystal structure determination of three heptapeptides containing alpha-aminoisobutyryl (Aib) residues as a means of helix stabilization provides a high-resolution characterization of 6-->1 hydrogen-bonded conformations, reminiscent of helix-terminating structural features in proteins. The crystal parameters for the three peptides, Boc-Val-Aib-X-Aib-Ala-Aib-Y-OMe, where X and Y are Phe, Leu (I), Leu, Phe (II) and Leu, Leu (III) are: (I) space group P1, Z = 1, a = 9.903 A, b = 10.709 A, c = 11.969 A, alpha = 102.94 degrees, beta = 103.41 degrees, gamma = 92.72 degrees, R = 4.55%; (II) space group P21, Z = 2, a = 10.052 A, b = 17.653 A, c = 13.510 A, beta = 108.45 degrees, R = 4.49%; (III) space group P1, Z = 2 (two independent molecules IIIa and IIIb in the asymmetric unit), a = 10.833 A, b = 13.850 A, c = 16.928 A, alpha = 99.77 degrees, beta = 105.90 degrees, gamma = 90.64 degrees, R = 8.54%. In all cases the helices form 3(10)/alpha-helical (or 3(10)helical) structures, with helical columns formed by head-to-tail hydrogen bonding. The helices assemble in an all-parallel motif in crystals I and III and in an antiparallel motif in II. In the four crystallographically characterized molecules, I, II, IIIa and IIIb, Aib(6) adopts a left-handed helical (hL) conformation with positive phi, psi values, resulting in 6-->1 hydrogen-bond formation between Aib(2) CO and Leu(7)/Phe(7) NH groups. In addition a 4-->1 hydrogen bond is seen between Aib(3) CO and Aib(6) NH groups. This pattern of hydrogen bonding is often observed at the C-terminus of helices proteins, with the terminal pi-type turn being formed by four residues adopting the hRhRhRhL conformation.

Aqueous channels within apolar peptide aggregates. Solvated helix of Boc-Aib-Ala-Leu-Aib-Ala-Leu-Aib- -Ala-Leu-Aib-OMe. H2O.CH3OH

Although the peptide Boc-Aibl-Ala2-Leu3- Aib4-Alas Leu'-Aib7-Ala8-Leu9-Aib'0-OMe [with a t-butoxycarbonyl(Boc) blocking group at the amino terminus, a methyl ester (OMe) at the carboxyl terminus, and four a-aminoisobutyric (Aib) residues] has a 3-fold repeat of residues, the helix formed by the peptide backbone is irregular. The carboxyl-terminal half assumes an at-helical form with torsion angles ) and r of approximately -60° and -45°, respectively, whereas the amino-terminal half is distorted by an insertion of a water molecule between the amide nitrogen of Ala5 [N(5)] and the carbonyl oxygen of Ala2 [0(2)]. The water molecule W(1) acts as a bridge by forming hydrogen bonds N(5).W(1) (2.93 A) and W(1)---0(2) (2.86 A). The distortion of the helix exposes the carbonyl oxygens of Aib' and Aib4 to the outside environment, with the consequence that the helix assumes an amphiphilic character despite having all apolar residues. Neighboring helices in the crystal run in antiparallel directions. On one side of a helix there are only hydrophobic contacts with efficient interdigitation of leucine side chains with those from the neighboring helix. On the other side of the helix there are hydrogen bonds between protruding carbonyl oxygens and four water molecules that separate two neighboring helices. Along the helix axis the helices bind head-to-tail with a direct hydrogen bond N(2)-0(9) (3.00 A). Crystals grown from methanol/water solution are in space group P2, with a = 15.778 ± 0.004 A, b = 11.228 ± 0.002 A, c = 18.415 ± 0.003 A, = 102.10 ± 0.02ur and two formula units per cell for C49HON1003 2H2OCH3OH. The overall agreement factorR is 7.5% for 3394 reflections observed with intensities >3a(F), and the resolution is 0.90 A.

A Study on the Success Potential of Multiple Mobile Payment Technologies

Payment systems all over the world have grown into a complicated web of solutions. This is more challenging in the case of mobile based payment systems. Mobile based payment systems are many and consist of different technologies providing different services. The diffusion of these various technologies in a market is uncertain. Diffusion theorists, for example Rogers, and Davis suggest how innovation is accepted in markets. In the case of electronic payment systems, the tale of Mondex vs Octopus throws interesting insights on diffusion. Our paper attempts to understand the success potential of various mobile payment technologies. We illustrate what we describe as technology breadth in mobile payment systems using data from payment systems all over the world (n=62). Our data shows an unexpected superiority of SMS technology, over other technologies like NFC, WAP and others. We also used a Delphi based survey (n=5) with experts to address the possibility that SMS will gain superiority in market diffusion. The economic conditions of a country, particularly in developing countries, the services availed and characteristics of the user (for example number of un-banked users in large populated countries) may put SMS in the forefront. This may be true more for micro payments using the mobile.

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.

Efficient algorithms for learning kernels from multiple similarity matrices with general convex loss functions

In this paper we consider the problem of learning an n × n kernel matrix from m(1) similarity matrices under general convex loss. Past research have extensively studied the m = 1 case and have derived several algorithms which require sophisticated techniques like ACCP, SOCP, etc. The existing algorithms do not apply if one uses arbitrary losses and often can not handle m > 1 case. We present several provably convergent iterative algorithms, where each iteration requires either an SVM or a Multiple Kernel Learning (MKL) solver for m > 1 case. One of the major contributions of the paper is to extend the well knownMirror Descent(MD) framework to handle Cartesian product of psd matrices. This novel extension leads to an algorithm, called EMKL, which solves the problem in O(m2 log n 2) iterations; in each iteration one solves an MKL involving m kernels and m eigen-decomposition of n × n matrices. By suitably defining a restriction on the objective function, a faster version of EMKL is proposed, called REKL,which avoids the eigen-decomposition. An alternative to both EMKL and REKL is also suggested which requires only an SVMsolver. Experimental results on real world protein data set involving several similarity matrices illustrate the efficacy of the proposed algorithms.

Robust formulations for clustering-based large-scale classification

Chebyshev-inequality-based convex relaxations of Chance-Constrained Programs (CCPs) are shown to be useful for learning classifiers on massive datasets. In particular, an algorithm that integrates efficient clustering procedures and CCP approaches for computing classifiers on large datasets is proposed. The key idea is to identify high density regions or clusters from individual class conditional densities and then use a CCP formulation to learn a classifier on the clusters. The CCP formulation ensures that most of the data points in a cluster are correctly classified by employing a Chebyshev-inequality-based convex relaxation. This relaxation is heavily dependent on the second-order statistics. However, this formulation and in general such relaxations that depend on the second-order moments are susceptible to moment estimation errors. One of the contributions of the paper is to propose several formulations that are robust to such errors. In particular a generic way of making such formulations robust to moment estimation errors is illustrated using two novel confidence sets. An important contribution is to show that when either of the confidence sets is employed, for the special case of a spherical normal distribution of clusters, the robust variant of the formulation can be posed as a second-order cone program. Empirical results show that the robust formulations achieve accuracies comparable to that with true moments, even when moment estimates are erroneous. Results also illustrate the benefits of employing the proposed methodology for robust classification of large-scale datasets.