Gravitational wave (GW) detection in space probes GW spectrum that is inaccessible from the Earth. In addition to LISA project led by European Space Agency, and the DECIGO detector proposed by the Japan Aerospace Exploration Agency, two Chinese space
-based GW observatories -- TianQin and Taiji -- are planned to be launched in the 2030s. TianQin has a unique concept in its design with a geocentric orbit. Taijis design is similar to LISA, but is more ambitious with longer arm distance. Both facilities are complementary to LISA, considering that TianQin is sensitive to higher frequencies and Taiji probes similar frequencies but with higher sensitivity. In this Perspective we explain the concepts for both facilities and introduce the development milestones of TianQin and Taiji projects in testing extraordinary technologies to pave the way for future space-based GW detections. Considering that LISA, TianQin and Taiji have similar scientific goals, all are scheduled to be launched around the 2030s and will operate concurrently, we discuss possible collaborations among them to improve GW source localization and characterization.
We consider moderately interacting particle systems with singular interaction kernel and environmental noise. It is shown that the mollified empirical measures converge in strong norms to the unique (local) solutions of nonlinear Fokker-Planck equati
ons. The approach works for the Biot-Savart and Poisson kernels.
We consider the problem of regularization by noise for the three dimensional magnetohydrodynamical (3D MHD) equations. It is shown that, in a suitable scaling limit, multiplicative noise of transport type gives rise to bounds on the vorticity fields
of the fluid velocity and magnetic fields. As a result, if the noise intensity is big enough, then the stochastic 3D MHD equations admit a pathwise unique global solution for large initial data, with high probability.
We consider on the torus the scaling limit of stochastic 2D (inviscid) fluid dynamical equations with transport noise to deterministic viscous equations. Quantitative estimates on the convergence rates are provided by combining analytic and probabili
stic arguments, especially heat kernel properties and maximal estimates for stochastic convolutions. Similar ideas are applied to the stochastic 2D Keller-Segel model, yielding explicit choice of noise to ensure that the blow-up probability is less than any given threshold. Our approach also gives rise to some mixing property for stochastic linear transport equations and dissipation enhancement in the viscous case.
We prove the existence of an eddy heat diffusion coefficient coming from an idealized model of turbulent fluid. A difficulty lies in the presence of a boundary, with also turbulent mixing and the eddy diffusion coefficient going to zero at the boundary. Nevertheless enhanced diffusion takes place.
We consider point vortex systems on the two dimensional torus perturbed by environmental noise. It is shown that, under a suitable scaling of the noises, weak limit points of the empirical measures are solutions to the vorticity formulation of deterministic 2D Navier-Stokes equations.
In this paper, we consider the following Kirchhoff type equation $$ -left(a+ bint_{R^3}| abla u|^2right)triangle {u}+V(x)u=f(u),,,xinR^3, $$ where $a,b>0$ and $fin C(R,R)$, and the potential $Vin C^1(R^3,R)$ is positive, bounded and satisfies suitabl
e decay assumptions. By using a new perturbation approach together with a new version of global compactness lemma of Kirchhoff type, we prove the existence and multiplicity of bound state solutions for the above problem with a general nonlinearity. We especially point out that neither the corresponding Ambrosetti-Rabinowitz condition nor any monotonicity assumption is required for $f$. Moreover, the potential $V$ may not be radially symmetry or coercive. As a prototype, the nonlinear term involves the power-type nonlinearity $f(u) = |u|^{p-2}u$ for $pin (2, 6)$. In particular, our results generalize and improve the results by Li and Ye (J.Differential Equations, 257(2014): 566-600), in the sense that the case $pin(2,3]$ is left open there.
Meta-learning for offline reinforcement learning (OMRL) is an understudied problem with tremendous potential impact by enabling RL algorithms in many real-world applications. A popular solution to the problem is to infer task identity as augmented st
ate using a context-based encoder, for which efficient learning of task representations remains an open challenge. In this work, we improve upon one of the SOTA OMRL algorithms, FOCAL, by incorporating intra-task attention mechanism and inter-task contrastive learning objectives for more effective task inference and learning of control. Theoretical analysis and experiments are presented to demonstrate the superior performance, efficiency and robustness of our end-to-end and model free method compared to prior algorithms across multiple meta-RL benchmarks.
Multiagent reinforcement learning (MARL) has achieved a remarkable amount of success in solving various types of video games. A cornerstone of this success is the auto-curriculum framework, which shapes the learning process by continually creating ne
w challenging tasks for agents to adapt to, thereby facilitating the acquisition of new skills. In order to extend MARL methods to real-world domains outside of video games, we envision in this blue sky paper that maintaining a diversity-aware auto-curriculum is critical for successful MARL applications. Specifically, we argue that emph{behavioural diversity} is a pivotal, yet under-explored, component for real-world multiagent learning systems, and that significant work remains in understanding how to design a diversity-aware auto-curriculum. We list four open challenges for auto-curriculum techniques, which we believe deserve more attention from this community. Towards validating our vision, we recommend modelling realistic interactive behaviours in autonomous driving as an important test bed, and recommend the SMARTS/ULTRA benchmark.
Pedestrian behavior prediction is one of the major challenges for intelligent driving systems. Pedestrians often exhibit complex behaviors influenced by various contextual elements. To address this problem, we propose BiPed, a multitask learning fram
ework that simultaneously predicts trajectories and actions of pedestrians by relying on multimodal data. Our method benefits from 1) a bifold encoding approach where different data modalities are processed independently allowing them to develop their own representations, and jointly to produce a representation for all modalities using shared parameters; 2) a novel interaction modeling technique that relies on categorical semantic parsing of the scenes to capture interactions between target pedestrians and their surroundings; and 3) a bifold prediction mechanism that uses both independent and shared decoding of multimodal representations. Using public pedestrian behavior benchmark datasets for driving, PIE and JAAD, we highlight the benefits of the proposed method for behavior prediction and show that our model achieves state-of-the-art performance and improves trajectory and action prediction by up to 22% and 9% respectively. We further investigate the contributions of the proposed reasoning techniques via extensive ablation studies.
