Hong Kong University of Science and Technology

About: Hong Kong University of Science and Technology is a education organization based out in Hong Kong, Hong Kong, China. It is known for research contribution in the topics: Computer science & Catalysis. The organization has 20126 authors who have published 52428 publications receiving 1965915 citations. The organization is also known as: HKUST & The Hong Kong University of Science and Technology.

Convolutional LSTM Network: a machine learning approach for precipitation nowcasting

Abstract: The goal of precipitation nowcasting is to predict the future rainfall intensity in a local region over a relatively short period of time. Very few previous studies have examined this crucial and challenging weather forecasting problem from the machine learning perspective. In this paper, we formulate precipitation nowcasting as a spatiotemporal sequence forecasting problem in which both the input and the prediction target are spatiotemporal sequences. By extending the fully connected LSTM (FC-LSTM) to have convolutional structures in both the input-to-state and state-to-state transitions, we propose the convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model for the precipitation nowcasting problem. Experiments show that our ConvLSTM network captures spatiotemporal correlations better and consistently outperforms FC-LSTM and the state-of-the-art operational ROVER algorithm for precipitation nowcasting.

Experiments and theory in strain gradient elasticity

Abstract: Conventional strain-based mechanics theory does not account for contributions from strain gradients. Failure to include strain gradient contributions can lead to underestimates of stresses and size-dependent behaviors in small-scale structures. In this paper, a new set of higher-order metrics is developed to characterize strain gradient behaviors. This set enables the application of the higher-order equilibrium conditions to strain gradient elasticity theory and reduces the number of independent elastic length scale parameters from five to three. On the basis of this new strain gradient theory, a strain gradient elastic bending theory for plane-strain beams is developed. Solutions for cantilever bending with a moment and line force applied at the free end are constructed based on the new higher-order bending theory. In classical bending theory, the normalized bending rigidity is independent of the length and thickness of the beam. In the solutions developed from the higher-order bending theory, the normalized higher-order bending rigidity has a new dependence on the thickness of the beam and on a higher-order bending parameter, bh. To determine the significance of the size dependence, we fabricated micron-sized beams and conducted bending tests using a nanoindenter. We found that the normalized beam rigidity exhibited an inverse squared dependence on the beam's thickness as predicted by the strain gradient elastic bending theory, and that the higher-order bending parameter, bh, is on the micron-scale. Potential errors from the experiments, model and fabrication were estimated and determined to be small relative to the observed increase in beam's bending rigidity. The present results indicate that the elastic strain gradient effect is significant in elastic deformation of small-scale structures.

Capital Structure Decisions: Which Factors Are Reliably Important?

Abstract: This paper examines the relative importance of many factors in the capital structure decisions of publicly traded American firms from 1950 to 2003. The most reliable factors for explaining market leverage are: median industry leverage (+ effect on leverage), market-to-book assets ratio (−), tangibility (+), profits (−), log of assets (+), and expected inflation (+). In addition, we find that dividend-paying firms tend to have lower leverage. When considering book leverage, somewhat similar effects are found. However, for book leverage, the impact of firm size, the market-to-book ratio, and the effect of inflation are not reliable. The empirical evidence seems reasonably consistent with some versions of the trade-off theory of capital structure.

Aggregation-Induced Emission: The Whole Is More Brilliant than the Parts

Abstract: "United we stand, divided we fall."--Aesop. Aggregation-induced emission (AIE) refers to a photophysical phenomenon shown by a group of luminogenic materials that are non-emissive when they are dissolved in good solvents as molecules but become highly luminescent when they are clustered in poor solvents or solid state as aggregates. In this Review we summarize the recent progresses made in the area of AIE research. We conduct mechanistic analyses of the AIE processes, unify the restriction of intramolecular motions (RIM) as the main cause for the AIE effects, and derive RIM-based molecular engineering strategies for the design of new AIE luminogens (AIEgens). Typical examples of the newly developed AIEgens and their high-tech applications as optoelectronic materials, chemical sensors and biomedical probes are presented and discussed.

VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator

Abstract: One camera and one low-cost inertial measurement unit (IMU) form a monocular visual-inertial system (VINS), which is the minimum sensor suite (in size, weight, and power) for the metric six degrees-of-freedom (DOF) state estimation. In this paper, we present VINS-Mono: a robust and versatile monocular visual-inertial state estimator. Our approach starts with a robust procedure for estimator initialization. A tightly coupled, nonlinear optimization-based method is used to obtain highly accurate visual-inertial odometry by fusing preintegrated IMU measurements and feature observations. A loop detection module, in combination with our tightly coupled formulation, enables relocalization with minimum computation. We additionally perform 4-DOF pose graph optimization to enforce the global consistency. Furthermore, the proposed system can reuse a map by saving and loading it in an efficient way. The current and previous maps can be merged together by the global pose graph optimization. We validate the performance of our system on public datasets and real-world experiments and compare against other state-of-the-art algorithms. We also perform an onboard closed-loop autonomous flight on the microaerial-vehicle platform and port the algorithm to an iOS-based demonstration. We highlight that the proposed work is a reliable, complete, and versatile system that is applicable for different applications that require high accuracy in localization. We open source our implementations for both PCs ( https://github.com/HKUST-Aerial-Robotics/VINS-Mono ) and iOS mobile devices ( https://github.com/HKUST-Aerial-Robotics/VINS-Mobile ).