arXiv Science⌕ Search

arXiv subjects

Fan Wang

Publications and source records attributed to Fan Wang.

At least 379 records · Page 21Linked to original sources
arXiv2019

Know More about Each Other: Evolving Dialogue Strategy via Compound Assessment

In this paper, a novel Generation-Evaluation framework is developed for multi-turn conversations with the objective of letting both participants know more about each other. For the sake of rational knowledge utilization and coherent conversation flow, a dialogue strategy which controls knowledge selection is instantiated and continuously adapted via reinforcement learning. Under the deployed strategy, knowledge grounded conversations are conducted with two dialogue agents. The generated dialogues are comprehensively evaluated on aspects like informativeness and coherence, which are aligned with our objective and human instinct. These assessments are integrated as a compound reward to guide the evolution of dialogue strategy via policy gradient. Comprehensive experiments have been carried out on the publicly available dataset, demonstrating that the proposed method outperforms the other state-of-the-art approaches significantly.

cs.CL↗
arXiv2019

Learning to Select Knowledge for Response Generation in Dialog Systems

End-to-end neural models for intelligent dialogue systems suffer from the problem of generating uninformative responses. Various methods were proposed to generate more informative responses by leveraging external knowledge. However, few previous work has focused on selecting appropriate knowledge in the learning process. The inappropriate selection of knowledge could prohibit the model from learning to make full use of the knowledge. Motivated by this, we propose an end-to-end neural model which employs a novel knowledge selection mechanism where both prior and posterior distributions over knowledge are used to facilitate knowledge selection. Specifically, a posterior distribution over knowledge is inferred from both utterances and responses, and it ensures the appropriate selection of knowledge during the training process. Meanwhile, a prior distribution, which is inferred from utterances only, is used to approximate the posterior distribution so that appropriate knowledge can be selected even without responses during the inference process. Compared with the previous work, our model can better incorporate appropriate knowledge in response generation. Experiments on both automatic and human evaluation verify the superiority of our model over previous baselines.

cs.CL↗
arXiv2019

Thermal enhanced NIR-NIR anti-Stokes emission in rare earth doped nanocrystals

Nanoparticles with anti-Stokes emissions enable many sensing applications, but their efficiencies are considerably low. The key to enable the process of anti-Stokes emissions is to create phonons and assist the excited photons to be pumped from a lower energy state onto a higher one. Increasing the temperature will generate more phonons, but it unavoidably quenches the luminescence. Here by quantifying the number of phonons being generated from the host crystal and at the surface of Yb3+/Nd3+ co-doped nanoparticles, we systematically investigated mechanisms towards the large enhancements of the phonon-assisted anti-Stokes emissions from 980 nm to 750 nm and 803 nm. Moreover, we provided direct evidence that moisture release from the nanoparticle surface at high temperature was not a main reason. We further demonstrated that the brightness of 10 nm nanoparticles were enhanced by more than two orders of magnitude, standing in stark contrast to the thermal quenching effect.

physics.optics↗
arXiv2019

Artificial Intelligence for Prosthetics - challenge solutions

In the NeurIPS 2018 Artificial Intelligence for Prosthetics challenge, participants were tasked with building a controller for a musculoskeletal model with a goal of matching a given time-varying velocity vector. Top participants were invited to describe their algorithms. In this work, we describe the challenge and present thirteen solutions that used deep reinforcement learning approaches. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each team implemented different modifications of the known algorithms by, for example, dividing the task into subtasks, learning low-level control, or by incorporating expert knowledge and using imitation learning.

cs.LG↗
arXiv2018

Stable bin packing of non-convex 3D objects with a robot manipulator

Recent progress in the field of robotic manipulation has generated interest in fully automatic object packing in warehouses. This paper proposes a formulation of the packing problem that is tailored to the automated warehousing domain. Besides minimizing waste space inside a container, the problem requires stability of the object pile during packing and the feasibility of the robot motion executing the placement plans. To address this problem, a set of constraints are formulated, and a constructive packing pipeline is proposed to solve for these constraints. The pipeline is able to pack geometrically complex, non-convex objects with stability while satisfying robot constraints. In particular, a new 3D positioning heuristic called Heightmap-Minimization heuristic is proposed, and heightmaps are used to speed up the search. Experimental evaluation of the method is conducted with a realistic physical simulator on a dataset of scanned real-world items, demonstrating stable and high-quality packing plans compared with other 3D packing methods.

cs.RO↗
arXiv2018

Intervention Aided Reinforcement Learning for Safe and Practical Policy Optimization in Navigation

Combining deep neural networks with reinforcement learning has shown great potential in the next-generation intelligent control. However, there are challenges in terms of safety and cost in practical applications. In this paper, we propose the Intervention Aided Reinforcement Learning (IARL) framework, which utilizes human intervened robot-environment interaction to improve the policy. We used the Unmanned Aerial Vehicle (UAV) as the test platform. We built neural networks as our policy to map sensor readings to control signals on the UAV. Our experiment scenarios cover both simulation and reality. We show that our approach substantially reduces the human intervention and improves the performance in autonomous navigation, at the same time it ensures safety and keeps training cost acceptable.

cs.RO↗
arXiv2018

Looking for the gluon condensation signature in proton using the Earth limb gamma-ray spectra

A new type of gamma ray spectrum is predicted in a general hadronic framework if considering the gluon condensation effects in proton. The result presents the power-law with a sharp break in the gamma ray spectra at the TeV-band. We suggest to probe this GC-signature in the Earth limb gamma-ray spectra using the DArk Matter Particle Explorer and the CALorimetric Electron Telescope on orbit.

hep-ph↗
arXiv2018

The gluon condensation effects in the DAMPE cosmic ray spectrum of electrons and positrons

Gluons dominate the proton behavior at high energy collisions, they can be condensed at ultra high energy. The collisions of the accelerated high energy protons with interplanetary matter in cosmic rays will produce a huge number of secondary particles at the gluon condensate energy region, which break the primary power-law of cosmic rays. The above predictions seem to be consistent with the recent DAMPE data concerning the electron plus positron spectra. We find that the smoothly broken power-law at $\sim 0.9 TeV$ and $3\sim 4 TeV$ in the DAMPE data can be understood as the gluon condensation effects in proton.

astro-ph.HE↗
arXiv2018

Hidden charmed states and multibody color flux-tube dynamics

Within the framework of the color flux-tube model with a multibody confinement potential, we systematically investigate the hidden charmed states observed in recent years. It can be found that most of them can be described as the compact tetraquark states $[cq][\bar{c}\bar{q}]$ ($q=u,d$ and $s$) in the color flux-tube model. The multibody confinement potential based on the color flux-tube picture is a dynamical mechanism in the formation and decay of the compact tetraquark states.

hep-ph↗
arXiv2018

Possible existence of a dibaryon candidate $NΔ$ ($D_{21}$)

Inspired by the experimental report by WASA-at-COSY Collaboration, we investigate the possibile existence of the dibaryon candidate $NΔ$ with quantum numbers $IJ^P=21^+$ ($D_{21}$). The dynamical calculation shows that we cannot obtain the bound $D_{21}$ state in the models which can obtain the experimental $d^{*}$. %although the $D_{21}$ state can bound in the range of parameters in quark models. The low-energy scattering phase shifts of the $N-Δ$ scattering give the same conclusion. Besides, the mass calculation by using the Gursey-Radicati mass formula and the analysis of the matrix elements of the color magnetic interaction show that the mass of $D_{21}$ is larger than that of $D_{12}$ ($NΔ$ with $IJ^P=12^+$), which indicate that it is less possible for the $D_{21}$ than the $D_{12}$ to form bound state.

hep-ph↗
arXiv2018

Geometric effects resulting from square and circular confinements for a particle constrained to a space curve

Investigating the geometric effects resulting from the detailed behaviors of the confining potential, we consider square and circular confinements to constrain a particle to a space curve. We find a torsion-induced geometric potential and a curvature-induced geometric momentum just in the square case, while a geometric gauge potential solely in the circular case. In the presence of electromagnetic field, a geometrically induced magnetic moment couples with magnetic field as an induced Zeeman coupling only for the circular confinement, also. As spin-orbit interaction is considered, we find some additional terms for the spin-orbit coupling, which are induced not only by torsion, but also curvature. Moreover, in the circular case, the spin also couples with an intrinsic angular momentum, which describes the azimuthal motions mapped on the space curve. As an important conclusion for the thin-layer quantization approach, some substantial geometric effects result from the confinement boundaries. Finally, these results are proved on a helical wire.

quant-ph↗
arXiv2018

Well-posedness of the free boundary problem in incompressible elastodynamics under the mixed type stability condition

We consider the free boundary problem for the incompressible elastodynamics equations. At the free boundary moving with the velocity of the fluid particles the columns of the deformation gradient are tangent to the boundary and the pressure vanishes outside the flow domain. We prove the local existence of a unique smooth solution of the free boundary problem, under the mixed type stability condition that some regions of the initial free boundary satisfy the Rayleigh-Taylor sign condition, while the remaining boundary satisfy the non-collinearity condition. In particular, we solve an open problem proposed by Y. Trakhinin in the paper \cite{Trak_18}.

math.AP↗
arXiv2018

Electromagnetic wave propagating along a space curve

Using the thin-layer approach, we derive the effective equation for the electromagnetic wave propagating along a space curve. We find intrinsic spin-orbit, extrinsic spin-orbit and extrinsic orbital angular momentum and intrinsic orbital angular momentum couplings induced by torsion, which can lead to geometric phase, spin and orbital Hall effects. And we show the helicity inversion induced by curvature that can convert the right-handed circularly polarized electromagnetic wave into left-handed polarized one, vice verse. Finally, we demonstrate that the gauge invariance of the effective dynamics is protected by the geometrically induced gauge potential.

physics.optics↗
arXiv2018

Search for a hidden strange baryon-meson bound state from $ϕ$ production in a nuclear medium

We investigate the hidden strange light baryon-meson system. With the resonating-group method, two bound states, $η'-N$ and $ϕ-N$, are found in the quark delocalization color screening model. Focusing on the $ϕ-N$ bound state around 1950\,MeV, we obtain the total decay width of about 4\,MeV by calculating the phase shifts in the resonance scattering processes. To study the feasibility of an experimental search for the $ϕ-N$ bound state, we perform a Monte Carlo simulation of the bound state production with an electron beam and a gold target. In the simulation, we use the CLAS12 detector with the Forward Tagger and the BONUS12 detector in Hall B at Jefferson Lab. Both the signal and the background channels are estimated. We demonstrate that the signal events can be separated from the background with some momentum cuts. Therefore it is feasible to experimentally search for the $ϕ-N$ bound state through the near threshold $ϕ$ meson production from heavy nuclei.

hep-ph↗
arXiv2017

Super-resolution imaging of quantum emitters in layered materials

Layered van der Waals materials are emerging as compelling two-dimensional (2D) platforms for studies of nanophotonics, polaritonics, valleytronics and spintronics, and have the potential to transform applications in sensing, imaging and quantum information processing. Amongst these, hexagonal boron nitride (hBN) is unique in that it hosts ultra-bright, room temperature single photon emitters (SPEs). However, an outstanding challenge is to locate SPEs in hBN with high precision, a task which requires breaking the optical diffraction limit. Here, we report the imaging of SPEs in layered hBN with a spatial resolution of 63 nm using ground state depletion (GSD) nanoscopy. Furthermore, we show that SPEs in hBN possess nonlinear photophysical properties which can be used to realize a new variant of GSD that employs a coincident pair of doughnut-shaped lasers to reduce the laser power that is needed to achieve a given resolution target. Our findings expand the current understanding of the photophysics of quantum emitters in layered hBN and demonstrate the potential for advanced nanophotonic and bio-imaging applications which require localization of individual emitters with super-resolution accuracy.

physics.optics↗
arXiv2017

Investigating the excited $Ω^{0}_{c}$ states through $Ξ_{c}K$ and $Ξ^{'}_{c}K$ decay channels

Inspired by the five newly observed $Ω^{0}_{c}$ states by the LHCb detector, we study the $Ω_{c}^{0}$ states as the $S-$wave molecular pentaquarks with $I=0$, $J^{P}=\frac{1}{2}^{-}$, $\frac{3}{2}^{-}$, and $\frac{5}{2}^{-}$ by solving the RGM equation in the framework of chiral quark model. Both the energies and the decay widths are obtained in this work. Our results suggest that $Ω_{c}(3119)^{0}$ can be explained as an $S-$wave resonance state of $ΞD$ with $J^{P}=\frac{1}{2}^{-}$, and the decay channels are the $S-$wave $Ξ_{c} K$ and $Ξ^{'}_{c}K$ . Other reported $Ω^{0}_{c}$ states cannot be obtained in our present calculation. Another $Ω_{c}^{0}$ state with much higher mass 3533 MeV with $J^{P}=\frac{5}{2}^{-}$ is also obtained. In addition, the calculation is extended to the $Ω_{b}^{0}$ states, similar results as that of $Ω^{0}_{c}$ are obtained.

hep-ph↗
arXiv2017

Seeds Cleansing CNMF for Spatiotemporal Neural Signals Extraction of Miniscope Imaging Data

Miniscope calcium imaging is increasingly being used to monitor large populations of neuronal activities in freely behaving animals. However, due to the high background and low signal-to-noise ratio of the single-photon based imaging used in this technique, extraction of neural signals from the large numbers of imaged cells automatically has remained challenging. Here we describe a highly accurate framework for automatically identifying activated neurons and extracting calcium signals from the miniscope imaging data, seeds cleansing Constrained Nonnegative Matrix Factorization (sc-CNMF). This sc-CNMF extends the conventional CNMF with two new modules: i) a neural enhancing module to overcome miniscope-specific limitations, and ii) a seeds cleansing module combining LSTM to rigorously select and cleanse the set of seeds for detecting regions-of-interest. Our sc-CNMF yields highly stable and superior performance in analyzing miniscope calcium imaging data compared to existing methods.

q-bio.NC↗