Multi-label text classification (MLTC) aims to tag most relevant labels for the given document. In this paper, we propose a Label-Specific Attention Network (LSAN) to learn the new document representation. LSAN takes advantage of label semantic information to determine the semantic connection between labels and document for constructing labelspecific document representation. Meanwhile, the self-attention mechanism is adopted to identify the label-specific document representation from document content information. In order to seamlessly integrate the above two parts, an adaptive fusion strategy is designed, which can effectively output the comprehensive document representation to build multilabel text classifier. Extensive experimental results on four benchmark datasets demonstrate that LSAN consistently outperforms the stateof-the-art methods, especially on the prediction of low-frequency labels. The code and hyper-parameter settings are released to facilitate other researchers 1 .
Permafrost has significant impacts on climate change through its strong interaction with the climate system. In order to better understand the permafrost variation and the role it plays in climate change, model outputs from Phase 5 of the Coupled Model Intercomparison Project (CMIP5) are used in the present study to diagnose the near‐surface permafrost on the Tibetan Plateau (TP), assess the abilities of the models to simulate present‐day (1986–2005) permafrost and project future permafrost change on the TP under four different representative concentration pathways (RCPs). The results indicate that estimations of present‐day permafrost using the surface frost index (SFI) and the Kudryavtsev method (KUD) show a spatial distribution similar to that of the frozen soil map on the TP. However, the permafrost area calculated via the KUD is larger than that calculated via the SFI. The SFI produces a present‐day permafrost area of 127.2 × 104 km2. The results also indicate that the permafrost on the TP will undergo regional degradation, mainly at the eastern, southern and northeastern edges, during the 21st century. Furthermore, most of the sustainable permafrost will probably exist only in the northwestern TP by 2099. The SFI also indicates that the permafrost area will shrink by 13.3 × 104 km2 (9.7%) and 14.6 × 104 km2 (10.5%) under the RCP4.5 and RCP8.5 scenarios, respectively, in the next 20 years and by 36.7 × 104 km2 (26.6%) and 45.7 × 104 km2 (32.7%), respectively, in the next 50 years. The results are helpful for us to better understand the permafrost response to climate change over the TP, further investigate the physical mechanism of the freeze–thaw process and improve the model parameterization scheme.
The energy management (EM) and driving speed co-optimization of a series hybrid electric vehicle (S-HEV) for minimizing fuel consumption is addressed in this article on the basis of a suitably modeled series powertrain architecture. The paper proposes a novel strategy that finds the optimal driving speed simultaneously with the energy source power split for the drive mission specified in terms of the road geometry and travel time. Such a combined optimization task is formulated as an optimal control problem that is solved by an indirect optimal control method, based on Pontryagin's minimum principle. The optimization scheme is tested under a rural drive mission by extensive comparisons with conventional methods that deal with either speed optimization only or EM strategies with given driving cycles. The comparative results show the superior performance of the proposed method and provide further insight into efficient driving.
Different from the traditional classification tasks which assume mutual exclusion of labels, hierarchical multi-label classification (HMLC) aims to assign multiple labels to every instance with the labels organized under hierarchical relations. Besides the labels, since linguistic ontologies are intrinsic hierarchies, the conceptual relations between words can also form hierarchical structures. Thus it can be a challenge to learn mappings from word hierarchies to label hierarchies. We propose to model the word and label hierarchies by embedding them jointly in the hyperbolic space. The main reason is that the tree-likeness of the hyperbolic space matches the complexity of symbolic data with hierarchical structures. A new Hyperbolic Interaction Model (HyperIM) is designed to learn the label-aware document representations and make predictions for HMLC. Extensive experiments are conducted on three benchmark datasets. The results have demonstrated that the new model can realistically capture the complex data structures and further improve the performance for HMLC comparing with the state-of-the-art methods. To facilitate future research, our code is publicly available.
A novel finite-time convergent estimation technique is proposed for identifying the amplitude, frequency and phase of a biased sinusoidal signal. Resorting to Volterra integral operators with suitably designed kernels, the measured signal is processed yielding a set of auxiliary signals in which the influence of the unknown initial conditions is removed. A second-order sliding mode-based adaptation law-fed by the aforementioned auxiliary signals-is designed for finite-time estimation of the frequency, amplitude, and phase. The worst case behavior of the proposed algorithm in presence of the bounded additive disturbances is fully characterized by Input-to-State Stability arguments. The effectiveness of the estimation technique is evaluated and compared with other existing tools via extensive numerical simulations.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.