Local imperceptible adversarial attacks against human pose estimation networks

Deep neural networks are vulnerable to attacks from adversarial inputs.Corresponding attack research on human pose estimation(HPE),particularly for body joint detection,has been largely unexplored.Transferring classification-based attack methods to body joint regression tasks is not straightforward.Another issue is that the attack effectiveness and imperceptibility contradict each other.To solve these issues,we propose local imperceptible attacks on HPE networks.In particular,we reformulate imperceptible attacks on body joint regression into a constrained maximum allowable attack.Furthermore,we approximate the solution using iterative gradient-based strength refinement and greedy-based pixel selection.Our method crafts effective perceptual adversarial attacks that consider both human perception and attack effectiveness.We conducted a series of imperceptible attacks against state-of-the-art HPE methods,including HigherHRNet,DEKR,and ViTPose.The experimental results demonstrate that the proposed method achieves excellent imperceptibility while maintaining attack effectiveness by significantly reducing the number of perturbed pixels.Approximately 4%of the pixels can achieve sufficient attacks on HPE.

Anomalous Cell Detection with Kernel Density-Based Local Outlier Factor

Since data services are penetrating into our daily life rapidly, the mobile network becomes more complicated, and the amount of data transmission is more and more increasing. In this case, the traditional statistical methods for anomalous cell detection cannot adapt to the evolution of networks, and data mining becomes the mainstream. In this paper, we propose a novel kernel density-based local outlier factor(KLOF) to assign a degree of being an outlier to each object. Firstly, the notion of KLOF is introduced, which captures exactly the relative degree of isolation. Then, by analyzing its properties, including the tightness of upper and lower bounds, sensitivity of density perturbation, we find that KLOF is much greater than 1 for outliers. Lastly, KLOFis applied on a real-world dataset to detect anomalous cells with abnormal key performance indicators(KPIs) to verify its reliability. The experiment shows that KLOF can find outliers efficiently. It can be a guideline for the operators to perform faster and more efficient trouble shooting.

Algorithm based on local breeding of growing modes for convection-allowing ensemble forecasting

We propose a method based on the local breeding of growing modes(LBGM) considering strong local weather characteristics for convection-allowing ensemble forecasting. The impact radius was introduced in the breeding of growing modes to develop the LBGM method. In the local breeding process, the ratio between the root mean square error(RMSE) of local space forecast at each grid point and that of the initial full-field forecast is computed to rescale perturbations. Preliminary evaluations of the method based on a nature run were performed in terms of three aspects: perturbation structure, spread,and the RMSE of the forecast. The experimental results confirm that the local adaptability of perturbation schemes improves after rescaling by the LBGM method. For perturbation physical variables and some near-surface meteorological elements, the LBGM method could increase the spread and reduce the RMSE of forecast,improving the performance of the ensemble forecast system.In addition, different from those existing methods of global orthogonalization approach, this new initial-condition perturbation method takes into full consideration the local characteristics of the convective-scale weather system, thus making convectionallowing ensemble forecast more accurate.

LOCAL ANALYSIS OF THE FULLY DISCRETE LOCAL DISCONTINUOUS GALERKIN METHOD FOR THE TIME-DEPENDENT SINGULARLY PERTURBED PROBLEM

In this paper we consider the fully discrete local discontinuous Galerkin method, where the third order explicit Runge-Kutta time marching is coupled. For the one-dimensional time-dependent singularly perturbed problem with a boundary layer, we shall prove that the resulted scheme is not only of good behavior at the local stability, but also has the double-optimal local error estimate. It is to say, the convergence rate is optimal in both space and time, and the width of the cut-off subdomain is also nearly optimal, if the boundary condition at each intermediate stage is given in a proper way. Numerical experiments are also given.

Perfectly Matched Layer Method for Acoustic Scattering Problem by a Locally Perturbed Line with Impedance Boundary Condition

In this paper,we study the two-dimensional Helmholtz scattering problem by a locally perturbed line with impedance boundary condition.Different from the problem with Dirichlet boundary condition,the Green function of the Helmholtz equation with impedance boundary condition becomes very complicated and comprises surface waves along the locally perturbed line.A uniaxial perfectly matched layer(UPML)method is proposed to truncate the half plane into a bounded computational domain.The main contribution of this paper is to prove the well-posedness of the PML problem and the exponential convergence of the approximate solution to the exact solution as either the thickness or the medium parameter of PML increases.

Distinct potential aerosol masses under different scenarios of transport at a suburban site of Beijing

In order to evaluate the secondary aerosol formation potential at a suburban site of Beijing,in situ perturbation experiments in a potential aerosol mass（PAM） reactor were carried out in the winter of 2014.The variations of secondary aerosol formation as a function of time,OH exposure,and the concentrations of gas phase pollutants and particles were reported in this study.Two periods with distinct secondary aerosol formation potentials,marked as Period Ⅰ and Period Ⅱ,were identified during the observation.In Period Ⅰ,the secondary aerosol formation potential was high,and correlated well to the air pollutants,i.e.,SO2,NO2,and CO.The maximal secondary aerosol formation was observed with an aging time equivalent to about 3 days of atmospheric oxidation.In period Ⅱ,the secondary aerosol formation potential was low,with no obvious correlation with the air pollutants.Meanwhile,the aerosol mass decreased,instead of showing a peak,with increasing aging time.Backward trajectory analysis during the two periods confirmed that the air mass in Period Ⅰwas mainly from local sources,while it was attributed mostly to long distance transport in Period Ⅱ.The air lost its reactivity during the long transport and the particles became highly aged,resulting in a low secondary aerosol formation potential.Our experimental results indicated that the in situ measurement of the secondary aerosol formation potential could provide important information for evaluating the contributions of local emission and long distance transport to the aerosol pollution.