Showing papers by "Nanjing University published in 2006"
TL;DR: The empirical study suggests that some methods that have been believed to be effective in addressing the class imbalance problem may, in fact, only be effective on learning with imbalanced two-class data sets.
Abstract: This paper studies empirically the effect of sampling and threshold-moving in training cost-sensitive neural networks. Both oversampling and undersampling are considered. These techniques modify the distribution of the training data such that the costs of the examples are conveyed explicitly by the appearances of the examples. Threshold-moving tries to move the output threshold toward inexpensive classes such that examples with higher costs become harder to be misclassified. Moreover, hard-ensemble and soft-ensemble, i.e., the combination of above techniques via hard or soft voting schemes, are also tested. Twenty-one UCl data sets with three types of cost matrices and a real-world cost-sensitive data set are used in the empirical study. The results suggest that cost-sensitive learning with multiclass tasks is more difficult than with two-class tasks, and a higher degree of class imbalance may increase the difficulty. It also reveals that almost all the techniques are effective on two-class tasks, while most are ineffective and even may cause negative effect on multiclass tasks. Overall, threshold-moving and soft-ensemble are relatively good choices in training cost-sensitive neural networks. The empirical study also suggests that some methods that have been believed to be effective in addressing the class imbalance problem may, in fact, only be effective on learning with imbalanced two-class data sets.
1,227 citations
TL;DR: Applications to two real-world multilabel learning problems, i.e., functional genomics and text categorization, show that the performance of BP-MLL is superior to that of some well-established multILabel learning algorithms.
Abstract: In multilabel learning, each instance in the training set is associated with a set of labels and the task is to output a label set whose size is unknown a priori for each unseen instance. In this paper, this problem is addressed in the way that a neural network algorithm named BP-MLL, i.e., backpropagation for multilabel learning, is proposed. It is derived from the popular backpropagation algorithm through employing a novel error function capturing the characteristics of multilabel learning, i.e., the labels belonging to an instance should be ranked higher than those not belonging to that instance. Applications to two real-world multilabel learning problems, i.e., functional genomics and text categorization, show that the performance of BP-MLL is superior to that of some well-established multilabel learning algorithms
1,075 citations
TL;DR: This review focuses on new endophyte-related findings in biology and ecology, and also summarises the various metabolites isolated from endophytes.
857 citations
INAF1, Pennsylvania State University2, University of Milan3, Goddard Space Flight Center4, National Research Council5, Kavli Institute for Theoretical Physics6, International School for Advanced Studies7, Weizmann Institute of Science8, University of Nevada, Las Vegas9, Nanjing University10, Polaris Industries11, University of Leicester12, Sonoma State University13, Universities Space Research Association14, Los Alamos National Laboratory15
TL;DR: A supernova is caught in the act of exploding, directly observing the shock break-out, which indicates that the GRB progenitor was a Wolf–Rayet star.
Abstract: A link between long γ-ray bursts (GRBs) and supernovae has been established, but whether there is a similar relationship between the weaker and softer X-ray flashes and supernovae is unclear. GRB/XRF 060218, spotted by the Swift satellite on 18 February this year, may supply that missing link. In the first of four papers on this novel burster, Campana et al. report the sighting of the X-ray signature of a shock break-out, possible evidence of a supernova in progress. Pian et al. report the optical discovery of a type Ic supernova 2006aj associated with GRB/XRF 060218. Soderberg et al. report radio and X-ray observations that show that XRF 060218 is 100 times less energetic than, but of a type that is ten times more common than cosmological GRBs. Mazzali et al. modelled the spectra and light curve of SN 2006aj to show that it had a much smaller explosion energy and ejected much less mass than other GRB-supernovae, suggesting that it was produced by a star with a mass was only about 20 times that of the Sun, leaving behind a neutron star, rather than a black hole. Observations of the close gamma-ray burst GRB 060218 and its connection to supernova SN 2006aj reveal the break-out of a shock wave driven by a mildly relativistic shell into the dense wind surrounding the GRB progenitor. These observation catch a supernova in the act of exploding. Although the link between long γ-ray bursts (GRBs) and supernovae has been established1,2,3,4, hitherto there have been no observations of the beginning of a supernova explosion and its intimate link to a GRB. In particular, we do not know how the jet that defines a γ-ray burst emerges from the star's surface, nor how a GRB progenitor explodes. Here we report observations of the relatively nearby GRB 060218 (ref. 5) and its connection to supernova SN 2006aj (ref. 6). In addition to the classical non-thermal emission, GRB 060218 shows a thermal component in its X-ray spectrum, which cools and shifts into the optical/ultraviolet band as time passes. We interpret these features as arising from the break-out of a shock wave driven by a mildly relativistic shell into the dense wind surrounding the progenitor7. We have caught a supernova in the act of exploding, directly observing the shock break-out, which indicates that the GRB progenitor was a Wolf–Rayet star.
835 citations
TL;DR: This note investigates process fault accommodation in a class of nonlinear continuous-time systems using a new fault estimation module, based on an adaptive estimator, to compensate for the effect of the faults by stabilizing the closed-loop system.
Abstract: This note investigates process fault accommodation in a class of nonlinear continuous-time systems. A new fault estimation module, based on an adaptive estimator, is first proposed. The fault tolerant controller is constructed to compensate for the effect of the faults by stabilizing the closed-loop system. A flexible joint robotic example is given to illustrate the efficiency of the proposed approach
712 citations
TL;DR: Categorize and evaluate face recognition algorithms that rely heavily on the size and representative of training set, and the prominent algorithms are described and critically analyzed.
707 citations
TL;DR: In this article, single-phase Bi1−xNdxFeO3 (BNFOx) multiferroic ceramics were prepared to study the effects of Nd substitution on their crystal structure and ferroelectromagnetic behavior.
Abstract: Single-phase Bi1−xNdxFeO3 (BNFOx) (x=0–02) multiferroic ceramics were prepared to study the effects of Nd substitution on their crystal structure and ferroelectromagnetic behavior Rietveld refinement of x-ray diffraction data showed a continual transformation of crystal structure from the rhombohedral structure of BNFOx=0 (BiFeO3) to a triclinic structure in BNFOx=005–015 and a pseudotetragonal structure in BNFOx=0175–02 Ferroelectromagnetic measurements revealed the existence of ferroelectricity with remnant polarization of ∼9μC∕cm2 in BNFOx=0–0175, paraelectricity in BNFOx=02, and weak ferromagnetism with remnant magnetizations of 007–0227emu∕g in BNFOx=015–02 Magnetoelectric coupling was obvious in BNFOx=015–0175 near the Neel temperature of ∼380°C
468 citations
TL;DR: Investigation in two cool-season turfgrass species showed it was insufficient to protect the chloroplast from damage with increase in heat stress level, even if with high AsA and GSH contents to remove reactive oxygen species (ROS).
461 citations
Proceedings Article•
04 Dec 2006TL;DR: This paper formalizes multi-instance multi-label learning, where each training example is associated with not only multiple instances but also multiple class labels, and proposes the MIMLBOOST and MIMLSVM algorithms which achieve good performance in an application to scene classification.
Abstract: In this paper, we formalize multi-instance multi-label learning, where each training example is associated with not only multiple instances but also multiple class labels Such a problem can occur in many real-world tasks, eg an image usually contains multiple patches each of which can be described by a feature vector, and the image can belong to multiple categories since its semantics can be recognized in different ways We analyze the relationship between multi-instance multi-label learning and the learning frameworks of traditional supervised learning, multi-instance learning and multi-label learning Then, we propose the MIMLBOOST and MIMLSVM algorithms which achieve good performance in an application to scene classification
455 citations
TL;DR: The distance to the massive star–forming region W3OH in the Perseus spiral arm of the Milky Way is measured by triangulation, with Earth's orbit as one segment of a triangle, using the Very Long Baseline Array, resolving the long-standing problem that there is a discrepancy between different techniques used to determine distances.
Abstract: We have measured the distance to the massive star-forming region W3OH in the Perseus spiral arm of the Milky Way to be 195 +/- 004 kiloparsecs (586 x10(16) km) This distance was determined by triangulation, with Earth's orbit as one segment of a triangle, using the Very Long Baseline Array This resolves the long-standing problem that there is a discrepancy of a factor of 2 between different techniques used to determine distances The reason for the discrepancy is that this portion of the Perseus arm has anomalous motions The orientation of the anomalous motion agrees with spiral density-wave theory, but the magnitude of the motion is somewhat larger than most models predict
429 citations
TL;DR: This work shows that x-ray flares discovered in two short gamma-ray bursts last much longer than the previously proposed postmerger energy-release time scales and can be produced by differentially rotating, millisecond pulsars after the mergers of binary neutron stars.
Abstract: Recent observations support the suggestion that short-duration gamma-ray bursts are produced by compact star mergers. The x-ray flares discovered in two short gamma-ray bursts last much longer than the previously proposed postmerger energy-release time scales. Here, we show that they can be produced by differentially rotating, millisecond pulsars after the mergers of binary neutron stars. The differential rotation leads to windup of interior poloidal magnetic fields and the resulting toroidal fields are strong enough to float up and break through the stellar surface. Magnetic reconnection-driven explosive events then occur, leading to multiple x-ray flares minutes after the original gamma-ray burst.
TL;DR: Hydrothermal reaction of (l)-N-(4'-cyanobenzy)-(S)-proline with CdCl2 as a Lewis acid catalyst and NaN3 gives colorless block compound 1, in which 1 displays a complicated 3D framework that makes it, by definition, a high dielectric material.
Abstract: Hydrothermal reaction of (l)-N-(4'-cyanobenzy)-(S)-proline with CdCl2 as a Lewis acid catalyst and NaN3 gives colorless block compound 1, in which 1 displays a complicated 3D framework. Ferroelectric and dielectric property measurements reveal that 1 exhibits physical properties comparable to that of a typical ferroelectric compound with a dipole relaxation process and a dielectric constant of ca. 38.6 that makes it, by definition, a high dielectric material.
TL;DR: In this paper, a novel CO 2 capture phenomenon is observed by modifying as-prepared mesoporous silica SBA-15 (SBA-P) with tetraethylenepentamine (TEPA), not only conserving the energy and time required for removing the template, but also opening the way to utilizing the micelle for dispersing guest species.
Abstract: A novel CO 2 capture phenomenon is observed by modifying as-prepared mesoporous silica SBA-15 (SBA(P)) with tetraethylenepentamine (TEPA), not only conserving the energy and time required for removing the template, but also opening the way to utilizing the micelle for dispersing guest species. The TEPA species dispersed within the channels of SBA(P) are highly accessible to CO 2 molecules; moreover, the hydroxyl group of the poly(ethylene oxide)-poly(propylene oxide)-poly(ethylene oxide) (Pluronic P123) template is able to modify the interactions between CO 2 and the amine to enhance the adsorptive capacity of this system. The remarkably high adsorption capacity (173 mg g -1 ) of this mesoporous silica-amine composite suggests potential CO 2 trapping applications, especially at low CO 2 concentrations during prolonged cyclic operations.
TL;DR: In this article, the influence of the mixture of polypropylene fiber and lime on the engineering properties of a clay-ey soil was investigated and analyzed through scanning electron microscopy (SEM) analysis of the specimens after shearing.
TL;DR: In this paper, the authors presented new laser ablation-ICP-MS U-Pb zircon data for the igneous rocks from Northern Guangxi, which indicated a broad duration of magmatic activities (ca 35 million years), inconsistent with plume models that predict widespread magmatic eruption and emplacement within period of 1-5 million years.
TL;DR: In this paper, single-phase insulating Bi1−xLaxFeO3 (BLFOx, x=0.30) ceramics were prepared and an obvious phase transition from rhombohedral to orthorhombic phase was observed near x = 0.30.
Abstract: Single-phase, insulating Bi1−xLaxFeO3 (BLFOx, x=0.05, 0.10, 0.15, 0.20, 0.30, and 0.40) ceramics were prepared. An obvious phase transition from rhombohedral to orthorhombic phase was observed near x=0.30. It is found that the phase transition destructs the spin cycloid of BiFeO3 (BFO), and therefore, releases the locked magnetization and enhances magnetoelectric interaction. As a result, improved multiferroic properties of the BLFO0.30 ceramics with remnant polarization and magnetization (2Pr and 2Mr) of 22.4μC∕cm2 and 0.041emu∕g, respectively, were established.
TL;DR: It was found that the electrical conductivity of the PANI/Au composite hollow spheres was more than 3 times higher than that of the pure PANI hollow spheres.
Abstract: Polyaniline (PANI)/Au composite hollow spheres were successfully synthesized using polystyrene/sulfonated polystyrene core/shell gel particle templates. The PANI shell thickness and the number of Au nanoparticles decorating the PANI could be controlled effectively by adjusting the experimental conditions. The morphology, composition, and optical properties of the resulting products were characterized by scanning electron microscopy, transmission electron microscopy, X-ray diffraction, thermogravimetric analysis, X-ray photoelectron spectroscopy, Fourier transform infrared spectroscopy, and ultraviolet-visible absorption spectra. It was found that the electrical conductivity of the PANI/Au composite hollow spheres was more than 3 times higher than that of the pure PANI hollow spheres. Furthermore, PANI/Au composites were immobilized on the surface of a glassy carbon electrode (GCE) and applied to construct a sensor. The obtained PANI/Au-modified GCEs showed one pair of redox peaks and high catalytic activity for the oxidation of dopamine. The possible formation mechanism of the PANI/Au composite hollow spheres was also discussed.
18 Dec 2006
TL;DR: Experiments show that the proposed algorithms, BalanceCascade and EasyEnsemble, have better AUC scores than many existing class-imbalance learning methods and have approximately the same training time as that of under-sampling, which trains significantly faster than other methods.
Abstract: Under-sampling is a class-imbalance learning method which uses only a subset of major class examples and thus is very efficient. The main deficiency is that many major class examples are ignored. We propose two algorithms to overcome the deficiency. EasyEnsemble samples several subsets from the major class, trains a learner using each of them, and combines the outputs of those learners. BalanceCascade is similar to EasyEnsemble except that it removes correctly classified major class examples of trained learners from further consideration. Experiments show that both of the proposed algorithms have better AUC scores than many existing class-imbalance learning methods. Moreover, they have approximately the same training time as that of under-sampling, which trains significantly faster than other methods.
TL;DR: Genipin and/or chemically modified variants of genipin are useful research tools for studying biological processes thought to be controlled by UCP2 and represent lead compounds that comprise a starting point for the development of therapies aimed at treating beta cell dysfunction.
TL;DR: In this paper, the fabrication methods as well as optical and electrical characteristics of silicon carbide nanocrystals, nanowires, nanotubes, and nanosized films are reviewed.
TL;DR: Hydro(solvo)thermal reactions between a new flexible multicarboxylate ligand of 2,2',3,3'-oxydiphthalic acid and M(NO(3))(2) afford two novel homochiral helical coordination polymers, each of which discriminates only one kind of crystallographical nonequivalent metal atom.
Abstract: Hydro(solvo)thermal reactions between a new flexible multicarboxylate ligand of 2,2‘,3,3‘-oxydiphthalic acid (2,2‘,3,3‘-H4ODPA) and M(NO3)2·xH2O (M = Zn, x = 6; M = Cd, x = 4) in the presence of 4,4‘-bipyridine (bpy) afford two novel homochiral helical coordination polymers {[Zn2(2,2‘,3,3‘-ODPA)(bpy)(H2O)3]·(H2O)2 for 1 and [Cd2(2,2‘,3,3‘-ODPA)(bpy)(H2O)3]·(H2O)2 for 2}. Though having almost the same chemical formula, they have different space groups (P212121 for 1 and P21 for 2) and different bridging modes of the 2,2‘,3,3‘-ODPA ligand. Two kinds of homochiral helices (right-handed) are found in both 1 and 2, each of which discriminates only one kind of crystallographical nonequivalent metal atom. 1 has a 2D metal−organic framework and can be seen as the unity of two parallel homochiral Zn1 and Zn2 helices, in which the nodes are etheric oxygen atoms. In contrast, 2 has a 3D metal−organic framework and consists of two partially overlapped homochiral Cd1 and Cd2 helices in the two dimensions. Moreover, me...
TL;DR: Wang et al. as discussed by the authors detected temporal trends and frequency changes at three major stations of Yangtze River, i.e. Yichang, Hankou and Datong representing upper, middle and lower reaches, respectively, with the help of parametric t-test, Mann- Kendall analysis and wavelet transform methods.
23 Oct 2006
TL;DR: The AGES (AGing pattErn Subspace) method for automatic age estimation is proposed, which aims to model the aging pattern, which is defined as a sequence of personal aging face images, by learning a representative subspace.
Abstract: Age Specific Human-Computer Interaction (ASHCI) has vast potential applications in daily life. However, automatic age estimation technique is still underdeveloped. One of the main reasons is that the aging effects on human faces present several unique characteristics which make age estimation a challenging task that requires non-standard classification approaches. According to the speciality of the facial aging effects, this paper proposes the AGES (AGing pattErn Subspace) method for automatic age estimation. The basic idea is to model the aging pattern, which is defined as a sequence of personal aging face images, by learning a representative subspace. The proper aging pattern for an unseen face image is then determined by the projection in the subspace that can best reconstruct the face image, while the position of the face image in that aging pattern will indicate its age. The AGES method has shown encouraging performance in the comparative experiments either as an age estimator or as an age range estimator.
TL;DR: In this paper, the structural features of the Longmen Shan fold-thrust belt are presented, using both seismic profiles and field data, and the major contributions of this article is finding preexisting Paleozoic rift basins under the Cenozoic thin-skinned thrust belts, which represent a new potential hydrocarbon play.
Abstract: The Longmen Shan fold-thrust belt is one of the key regions of demonstrable Mesozoic–Cenozoic tectonic evolution in China, and the Sichuan Basin was the first natural-gas-producing area in China. In this article, the structural features of the Longmen Shan belt are presented, using both seismic profiles and field data. The complex structures of the northeast-trending Longmen Shan fold-thrust belt and its foreland in the western Sichuan Basin are formed by southeast-directed thrusting. Several eastward-verging, rootless thrust sheets and imbricates of Cambrian–Triassic rocks have been recognized in the northern Longmen Shan belt. Evidence suggests that the northern Longmen Shan belt experienced at least two major periods of deformation in the Late Triassic and Cenozoic. However, the southern Longmen Shan belt is represented by the basement-involved thrust structures and klippen, and its major periods of deformation were in the latest Cretaceous–early Cenozoic. Sedimentary features in the western Sichuan Basin reflect a two-phase flexural-loading history and illustrate that the Late Triassic foreland basin extends along the foredeep of the entire length of the Longmen Shan belt, but the uppermost Cretaceous–Paleogene rejuvenated foreland basin is restricted in the southern part of the western Sichuan Basin. Structural geometries suggest that prospective traps are mainly developed in the frontal zone of the Longmen Shan fold-thrust belt and in the southern part of the western Sichuan Basin. One of the major contributions of this article is finding preexisting Paleozoic rift basins under the Cenozoic thin-skinned thrust belts, which represent a new potential hydrocarbon play.
TL;DR: The highly sensitive H atom Rydberg tagging time-of-flight method is used to conduct a crossed molecular beam scattering study of the F + H2 → HF + H reaction with full quantum-state resolution, finding pronounced forward-scattered HF products in the v′ = 2 vibrational state.
Abstract: Reaction resonances, or transiently stabilized transition-state structures, have proven highly challenging to capture experimentally. Here, we used the highly sensitive H atom Rydberg tagging time-of-flight method to conduct a crossed molecular beam scattering study of the F + H2 --> HF + H reaction with full quantum-state resolution. Pronounced forward-scattered HF products in the v' = 2 vibrational state were clearly observed at a collision energy of 0.52 kcal/mol; this was attributed to both the ground and the first excited Feshbach resonances trapped in the peculiar HF(v' = 3)-H' vibrationally adiabatic potential, with substantial enhancement by constructive interference between the two resonances.
Proceedings Article•
16 Jul 2006TL;DR: It is advocated that before applying the rescaling approach, the consistency of the costs must be examined at first, and it is better to apply rescaling after decomposing the multi‐class problem into a series of two‐class problems.
Abstract: A popular approach to cost-sensitive learning is to rescale the classes according to their misclassification costs. Although this approach is effective in dealing with binary-class problems, recent studies show that it is often not so helpful when being applied to multi-class problems directly. This paper analyzes that why the traditional rescaling approach is often helpless on multi-class problems, which reveals that before applying rescaling, the consistency of the costs must be examined. Based on the analysis. a new approach is presented, which should be the choice if the user wants to use rescaling for multi-class cost-sensitive learning. Moreover, this paper shows that the proposed approach is helpful when unequal misclassification costs and class imbalance occur simultaneously, and can also be used to tackle pure class-imbalance learning. Thus, the preposed approach provides a unified framework for using rescaling to address multi-class cost-sensitive learning as well as multi-class class-imbalance learning.
TL;DR: In this article, four interesting cadmium-II or zinc-II metal coordination polymers, namely [Cd2(2,3-pydc)2(bix)3·2H2O]n (1), [Cc2(μ2-OH2)(2,6-pyridine-2, 3-dicarboxylic acid)2 (bix)]n (2), [cc3(SIP)2,SIP]4·8H 2O],n (3), and
Abstract: Four novel interesting cadmium(II) or zinc(II) metal coordination polymers, [Cd2(2,3-pydc)2(bix)3·2H2O]n (1), [Cd2(μ2-OH2)(2,6-pydc)2(bix)]n (2), [Cd3(SIP)2(bix)4·8H2O]n (3), and [Zn2(SIP)(bix)3(OH)·2H2O]n (4) (bix = 1,4-bis(imidazol-1-ylmethyl)-benzene; 2,3-pydc = pyridine-2,3-dicarboxylic acid; 2,6-pydc = pyridine-2,6-dicarboxylic acid; SIP = 5-sulfoisophthalic acid monosodium salt), have been synthesized under hydrothermal conditions and structurally characterized. Polymer 1 features a 3D porous framework with uncoordinated water molecules trapped in the pores. Polymer 2 is a 2D infinite layer framework, and the resulting 2D structure is interconnected by hydrogen-bond interactions to lead to a 3D supramolecular architecture. Polymer 3 also possesses a 3D porous framework; the most prominent cavities are parallel to the a- and b- directions and are filled with free water molecules. Polymer 4 exhibits 4-fold interpenetration related by three different translation vectors. These four compounds exhibit st...
TL;DR: A novel reliability-based stochastic user equilibrium traffic assignment model in view of the day-to-day demand fluctuations for multi-class transportation networks and the efficiency of solution algorithm is presented.
Abstract: This paper presents a novel reliability-based stochastic user equilibrium traffic assignment model in view of the day-to-day demand fluctuations for multi-class transportation networks. In the model, each class of travelers has a different safety margin for on-time arrival in response to the stochastic travel times raised from demand variations. Travelers' perception errors on travel time are also considered in the model. This model is formulated as an equivalent variational inequality problem, which is solved by the proposed heuristic solution algorithm. Numerical examples are presented to illustrate the applications of the proposed model and the efficiency of solution algorithm.
TL;DR: In this paper, a universal formation mechanism based on the growth, split, wrapping, and thickening of Na2Ti3O7 nanointermediates is proposed, which links all kinds of morphologies observed for H2TiO7 nanentities.
Abstract: The formation of H2Ti3O7 nanotubes, nanofibers, nanowires, and nanobelts via alkali hydrothermal synthesis was studied in detail by TEM and HRTEM. The effects of preparation parameters, such as reaction temperature, duration, and cooling process, on the morphologies of the products are clarified. A universal formation mechanism is proposed based on the growth, split, wrapping, and thickening of Na2Ti3O7 nanointermediates, which links all kinds of morphologies observed for H2Ti3O7 nanoentities.
TL;DR: This fossil demonstrates that some mammaliaforms, or proximal relatives to modern mammals, developed diverse locomotory and feeding adaptations and were ecomorphologically different from the majority of generalized small terrestrial Mesozoic mammalian insectivores.
Abstract: A docodontan mammaliaform from the Middle Jurassic of China possesses swimming and burrowing skeletal adaptations and some dental features for aquatic feeding. It is the most primitive taxon in the mammalian lineage known to have fur and has a broad, flattened, partly scaly tail analogous to that of modern beavers. We infer that docodontans were semiaquatic, convergent to the modern platypus and many Cenozoic placentals. This fossil demonstrates that some mammaliaforms, or proximal relatives to modern mammals, developed diverse locomotory and feeding adaptations and were ecomorphologically different from the majority of generalized small terrestrial Mesozoic mammalian insectivores.