A Novel Bidirectional LSTM and Attention Mechanism Based Neural Network for Answer Selection in Community Question Answering

Deep learning models have been shown to have great advantages in answer selection tasks.The existing models,which employ encoder-decoder recurrent neural network(RNN),have been demonstrated to be effective.However,the traditional RNN-based models still suffer from limitations such as 1)high-dimensional data representation in natural language processing and 2)biased attentive weights for subsequent words in traditional time series models.In this study,a new answer selection model is proposed based on the Bidirectional Long Short-Term Memory(Bi-LSTM)and attention mechanism.The proposed model is able to generate the more effective question-answer pair representation.Experiments on a question answering dataset that includes information from multiple fields show the great advantages of our proposed model.Specifically,we achieve a maximum improvement of 3.8%over the classical LSTM model in terms of mean average precision.

The Comparison of Biologic Characteristics between Mice Embryonic Stem Cells and Bone Marrow Derived Dendritic Cells

OBJECTIVE This research was to induce dendritic cells (DCs)from mice embryonic stem cells and bone marrow mononuclearcells in vitro, and then compare the biologic characteristics of them.METHODS Embryonic stem cells (ESCs) suspending culturedin petri dishes were induced to generate embryonic bodies (EBs).Fourteen-day well-developed EBs were transferred to histologicalculture with the same medium and supplemented 25 ng/ml GM-CSF and 25 ng/ml IL-3. In the next 2 weeks, there were numerousimmature DCs outgrown. Meantime, mononuclear cells isolatedfrom mice bone marrow were induced to derive dendritic cells bysupplementing 25 ng/ml GM-CSF and 25 ng/ml IL-4, and then themorphology, phenotype and function of both dendritic cells fromdifferent origins were examined.RESULTS Growing mature through exposure tolipopolysaccharide (LPS), both ESC-DCs and BM-DCs exhibiteddramatic veils of cytoplasm and extensive dendrites on theirsurfaces, highly expressed CD11c, MHC-Ⅱ and CD86 with strongcapacity to stimulate primary T cell responses in mixed leukocytereaction (MLR) .CONCLUSION ESC-DC has the same biologic characteristics asBM-DC, and it provides a new, reliable source for the functionalresearch of DC and next produce corresponding anti-tumorvaccine.

Design and application of an Electric Tail Rotor Drive Control(ETRDC) for helicopters with performance tests

With the development of electric helicopters’ motor technology and the widespread use of electric drive rotors, more aircraft use electric rotors to provide thrust and directional control.For a helicopter tail rotor, the wake of the main rotor influences the tail rotor’s inflow and wake.In the procedure of controlling, crosswind will also cause changes to the tail disk load. This paper describes requirements and design principles of an electric motor drive and variable pitch tail rotor system. A particular spoke-type architecture of the motor is designed, and the performance of blades is analyzed by the CFD method. The demand for simplicity of moving parts and strict constraints on the weight of a helicopter makes the design of electrical and mechanical components challenging. Different solutions have been investigated to propose an effective alternative to the mechanical actuation system. A test platform is constructed which can collect the dynamic response of the thrust control. The enhancement of the response speed due to an individual motor speed control and variable-pitch system is validated.

AI Enabled Wireless Communications with Real Channel Measurements:Channel Feedback

Artificial intelligence(AI)has shown great potential in wireless communications.AI-empowered communication algorithms have beaten many traditional algorithms through simulations.However,the existing works just use the simulated datasets to train and test the algorithms,which can not represent the power of AI in practical communication systems.Therefore,Peng Cheng Laboratory holds an AI competition,National Artificial Intelligence Competition(NAIC):AI+wireless communications,in which one of the topics is AI-empowered channel feedback system design using practical measurements.In this paper,we give a baseline neural network design,QuanCsiNet,for this competition,and the details of the channel measurements.QuanCsiNet shows excellent performance on channel feedback and the complexity of the neural networks is also given.

Acceleration of unsteady vortex lattice method via dipole panel fast multipole method

The Unsteady Vortex Lattice Method(UVLM) is a medium-fidelity aerodynamic tool that has been widely used in aeroelasticity and flight dynamics simulations. The most timeconsuming step is the evaluation of the induced velocity. Supposing that the number of bound and wake lattices is N and the computational cost is O (N2), we present an OeNT Dipole Panel Fast Multipole Method(DPFMM) for the rapid evaluation of the induced velocity in UVLM. The multipole expansion coefficients of a quadrilateral dipole panel have been derived in spherical coordinates, whose accuracy is the same as that of the Biot-Savart kernel at the same truncation degree P.Two methods(the loosening method and the shrinking method) are proposed and tested for space partitioning volumetric panels. Compared with FMM for vortex filaments(with three harmonics),DPFMM is approximately two times faster for N2 [103,106]. The simulation time of a multirotor(N~104) is reduced from 100 min(with unaccelerated direct solver) to 2 min(with DPFMM).