Search

Accurate and reliable prediction has profound implications to a wide range of applications. In this study, we focus on an instance of spatio-temporal learning problem--traffic prediction--to demonstrate an advanced deep learning model developed for making accurate and reliable forecast. Despite the significant progress in traffic prediction, limited studies have incorporated both explicit and implicit traffic patterns simultaneously to improve prediction performance. Meanwhile, the variability nature of traffic states necessitates quantifying the uncertainty of model predictions in a statistically principled way; however, extant studies offer no provable guarantee on the statistical validity of confidence intervals in reflecting its actual likelihood of containing the ground truth. In this paper, we propose an end-to-end traffic prediction framework that leverages three primary components to generate accurate and reliable traffic predictions: dynamic causal structure learning for discovering implicit traffic patterns from massive traffic data, causally-aware spatio-temporal multi-graph convolution network (CASTMGCN) for learning spatio-temporal dependencies, and conformal prediction for uncertainty quantification. CASTMGCN fuses several graphs that characterize different important aspects of traffic networks and an auxiliary graph that captures the effect of exogenous factors on the road network. On this basis, a conformal prediction approach tailored to spatio-temporal data is further developed for quantifying the uncertainty in node-wise traffic predictions over varying prediction horizons. Experimental results on two real-world traffic datasets demonstrate that the proposed method outperforms several state-of-the-art models in prediction accuracy; moreover, it generates more efficient prediction regions than other methods while strictly satisfying the statistical validity in coverage.

In this paper, we develop a generic methodology to encode hierarchical causality structure among observed variables into a neural network in order to improve its predictive performance. The proposed methodology, called causality-informed neural network (CINN), leverages three coherent steps to systematically map the structural causal knowledge into the layer-to-layer design of neural network while strictly preserving the orientation of every causal relationship. In the first step, CINN discovers causal relationships from observational data via directed acyclic graph (DAG) learning, where causal discovery is recast as a continuous optimization problem to avoid the combinatorial nature. In the second step, the discovered hierarchical causality structure among observed variables is systematically encoded into neural network through a dedicated architecture and customized loss function. By categorizing variables in the causal DAG as root, intermediate, and leaf nodes, the hierarchical causal DAG is translated into CINN with a one-to-one correspondence between nodes in the causal DAG and units in the CINN while maintaining the relative order among these nodes. Regarding the loss function, both intermediate and leaf nodes in the DAG graph are treated as target outputs during CINN training so as to drive co-learning of causal relationships among different types of nodes. As multiple loss components emerge in CINN, we leverage the projection of conflicting gradients to mitigate gradient interference among the multiple learning tasks. Computational experiments across a broad spectrum of UCI data sets demonstrate substantial advantages of CINN in predictive performance over other state-of-the-art methods. In addition, an ablation study underscores the value of integrating structural and quantitative causal knowledge in enhancing the neural network's predictive performance incrementally.

In this paper, we investigate the tensor form factors of $P\to P,\,S,\,V$ and $A$ transitions within the standard light-front (SLF) and the covariant light-front (CLF) quark models (QMs). The self-consistency and Lorentz covariance of CLF QM are analyzed via these quantities, and the effects of zero-mode are discussed. For the $P\to V$ and $A$ transitions, besides the inconsistence between the results extracted via longitudinal and transverse polarization states, which is caused by the residual $\omega$-dependent spurious contributions, we find and analyze a "new" self-consistence problem of the traditional CLF QM, which is caused by the different strategies for dealing deal with the trace term in CLF matrix element. A possible solution to the problems of traditional CLF QM is discussed and confirmed numerically. Finally, the theoretical predictions for the tensor form factors of some $c\to q,\,s$ and $b\to q,\,s\,,c$ ($q=u,d$) induced $P\to P,\,S,\,V$ and $A$ transitions are updated within the CLF QM with a self-consistent scheme., Comment: 34 pages, 9 figures and 7 tables. arXiv admin note: text overlap with arXiv:1908.04677

In this paper, we investigate the tree-dominated $\bar{B}^*_{u,d,s,c} \to V \ell^- \bar{\nu}_\ell$ ($V=D^*_{u,d}\,,D^*_s\,,J/\psi$ and $\ell=e\,,\mu\,,\tau$) decays in the Standard Model with the relevant form factors obtained in the light-front quark model. These decays involve much more helicity states relative to the corresponding $\bar{B}^* \to P \ell^- \bar{\nu}_\ell$ and $\bar{B} \to V \ell^- \bar{\nu}_\ell$ decays, and moreover, the contribution of longitudinal polarization mode ($V$ meson) is relatively small, $\sim 30\%$, compared with the corresponding $B$ meson decays. We have also computed the branching fraction, lepton spin asymmetry, forward-backward asymmetry and ratio $R_V^{\ast(L)}\equiv \frac{\mathcal{B}(\bar{B}^*\to V\tau^- \bar{\nu}_\tau)}{\mathcal{B}(\bar{B}^*\to V \ell^{\prime-} \bar{\nu}_{\ell^{\prime}})}$ ($\ell'=e\,,\mu$). Numerically, the branching fractions of $\bar{B}^* \to V \ell^{\prime-} \bar{\nu}_{\ell^{\prime}}$ decays are at the level of ${\cal O}(10^{-7})$, and are hopeful to be observed by LHC and Belle-II experiments. The ratios $R_{D^*\,,D^*_s\,,J/\psi}^{\ast(L)}$ have relatively small theoretical uncertainties and are close to each other, $R^{*(L)}_{D^*}\simeq R^{*(L)}_{D^*_s}\simeq R^{*(L)}_{J/\psi} \simeq [0.26,0.27]~([0.27,0.29])$, which are a bit different from the predictions in some previous works. The future measurements are expected to make tests on these predictions., Comment: 22 pages, 3 figures

We study the electronic structure of WTe$_2$ thin film fakes with different thickness down to 11 nm. Angle-dependent quantum oscillations reveal a crossover from a three-dimensional (3D) to a two-dimensional (2D) electronic system when the sample thickness is reduced below 26 nm. The quantum oscillations further show that the Fermi pockets get smaller as the samples are made thinner, indicating that the overlap between conduction and valence bands is getting smaller and implying the spatial confinement could lift the overlap in even thinner samples. In addition, the quadratic magnetoresistance (MR) also shows a crossover from 3D to 2D behavior as the samples are made thinner, while gating is shown to affect both the quadratic MR and the quantum oscillations of a thin sample by tuning its carrier density., Comment: 10 pages, 7 figures, supplementary material included

It is proposed that the new generation of spintronics should be ideally massless and dissipationless for the realization of ultra-fast and ultra-low-power spintronic devices. We demonstrate that the spin-gapless materials with linear energy dispersion are unique materials that can realize these massless and dissipationless states. Furthermore, we propose four new types of spin Hall effects which consist of spin accumulation of equal numbers of electrons and holes having the same or opposite spin polarization at the sample edge in Hall effect measurements, but with vanishing Hall voltage. These new Hall effects can be classified as (quantum) anomalous spin Hall effects. The physics for massless and dissipationless spintronics and the new spin Hall effects are presented for spin-gapless semiconductors with either linear or parabolic dispersion. New possible candidates for Dirac-type or parabolic type spin-gapless semiconductors are demonstrated in ferromagnetic monolayers of simple oxides with either honeycomb or square lattices., Comment: 5 pages, 7 figues

In anticipation of abundant $B^*$ data samples at high-luminosity heavy-flavor experiments in the future, the tree-dominated semileptonic $\bar{B}^*_{u,d,s} \to P \ell^- \bar{\nu}_\ell$ $(P=D\,,D_s\,,\pi\,,K)$ decays are studied within the Standard Model. After a detailed calculation of the helicity amplitudes, the theoretical predictions for branching fraction~(decay rate), lepton spin asymmetry, forward-backward asymmetry and ratio $R_D^{\ast(L)}$ are firstly presented. It is found that the CKM-favored $\bar{B}^* \to D \ell^- \bar{\nu}_\ell$ decays have relatively large branching fractions of ${\cal O}(10^{-9})$$\sim$${\cal O}(10^{-7})$, and are in the scope of running LHC and forthcoming SuperKEKB/Belle-II experiments., Comment: 18 pages, 2 figures