Search

In this paper, we introduce a divide-and-conquer algorithm to improve sentence alignment speed. We utilize external bilingual sentence embeddings to find accurate hard delimiters for the parallel texts to be aligned. We use Monte Carlo simulation to show experimentally that using this divide-and-conquer algorithm, we can turn any quadratic time complexity sentence alignment algorithm into an algorithm with average time complexity of O(NlogN). On a standard OCR-generated dataset, our method improves the Bleualign baseline by 3 F1 points. Besides, when computational resources are restricted, our algorithm is faster than Vecalign in practice.

Deep neural networks are vulnerable to adversarial examples, which can fool deep models by adding subtle perturbations. Although existing attacks have achieved promising results, it still leaves a long way to go for generating transferable adversarial examples under the black-box setting. To this end, this paper proposes to improve the transferability of adversarial examples, and applies dual-stage feature-level perturbations to an existing model to implicitly create a set of diverse models. Then these models are fused by the longitudinal ensemble during the iterations. The proposed method is termed Dual-Stage Network Erosion (DSNE). We conduct comprehensive experiments both on non-residual and residual networks, and obtain more transferable adversarial examples with the computational cost similar to the state-of-the-art method. In particular, for the residual networks, the transferability of the adversarial examples can be significantly improved by biasing the residual block information to the skip connections. Our work provides new insights into the architectural vulnerability of neural networks and presents new challenges to the robustness of neural networks.

An adversary can fool deep neural network object detectors by generating adversarial noises. Most of the existing works focus on learning local visible noises in an adversarial "patch" fashion. However, the 2D patch attached to a 3D object tends to suffer from an inevitable reduction in attack performance as the viewpoint changes. To remedy this issue, this work proposes the Coated Adversarial Camouflage (CAC) to attack the detectors in arbitrary viewpoints. Unlike the patch trained in the 2D space, our camouflage generated by a conceptually different training framework consists of 3D rendering and dense proposals attack. Specifically, we make the camouflage perform 3D spatial transformations according to the pose changes of the object. Based on the multi-view rendering results, the top-n proposals of the region proposal network are fixed, and all the classifications in the fixed dense proposals are attacked simultaneously to output errors. In addition, we build a virtual 3D scene to fairly and reproducibly evaluate different attacks. Extensive experiments demonstrate the superiority of CAC over the existing attacks, and it shows impressive performance both in the virtual scene and the real world. This poses a potential threat to the security-critical computer vision systems.

Catastrophic forgetting in continual learning is a common destructive phenomenon in gradient-based neural networks that learn sequential tasks, and it is much different from forgetting in humans, who can learn and accumulate knowledge throughout their whole lives. Catastrophic forgetting is the fatal shortcoming of a large decrease in performance on previous tasks when the model is learning a novel task. To alleviate this problem, the model should have the capacity to learn new knowledge and preserve learned knowledge. We propose an average gradient episodic memory (A-GEM) with a soft constraint $\epsilon \in [0, 1]$, which is a balance factor between learning new knowledge and preserving learned knowledge; our method is called gradient episodic memory with a soft constraint $\epsilon$ ($\epsilon$-SOFT-GEM). $\epsilon$-SOFT-GEM outperforms A-GEM and several continual learning benchmarks in a single training epoch; additionally, it has state-of-the-art average accuracy and efficiency for computation and memory, like A-GEM, and provides a better trade-off between the stability of preserving learned knowledge and the plasticity of learning new knowledge., Comment: 20 pages, 12 figures

Fast gradient sign attack series are popular methods that are used to generate adversarial examples. However, most of the approaches based on fast gradient sign attack series cannot balance the indistinguishability and transferability due to the limitations of the basic sign structure. To address this problem, we propose a method, called Adam Iterative Fast Gradient Tanh Method (AI-FGTM), to generate indistinguishable adversarial examples with high transferability. Besides, smaller kernels and dynamic step size are also applied to generate adversarial examples for further increasing the attack success rates. Extensive experiments on an ImageNet-compatible dataset show that our method generates more indistinguishable adversarial examples and achieves higher attack success rates without extra running time and resource. Our best transfer-based attack NI-TI-DI-AITM can fool six classic defense models with an average success rate of 89.3% and three advanced defense models with an average success rate of 82.7%, which are higher than the state-of-the-art gradient-based attacks. Additionally, our method can also reduce nearly 20% mean perturbation. We expect that our method will serve as a new baseline for generating adversarial examples with better transferability and indistinguishability., Comment: Accepted to AAAI2022

Tiny suspensions of solid particles or liquid droplets, called aerosols, hover in earth's atmosphere and can be found over just about anywhere including oceans, deserts, vegetated areas, and other global regions. Aerosols come in a variety of sizes, shapes, and compositions which depend on such factors as their origin and how long they have been in the atmosphere (i.e., their residence time). Some of the more common types of aerosols include mineral dust and sea salt which get lifted from the desert and ocean surfaces, respectively by mechanical forces such as strong winds. Depending on their size, aerosols will either fall out gravitationally, as in the case of larger particles, or will remain resident in the atmosphere where they can undergo further change through interactions with other aerosols and cloud particles. Not only do aerosols affect air quality where they pose a health risk, they can also perturb the distribution of radiation in the earth-atmosphere system which can inevitably lead to changes in our climate. One aerosol that has been in the forefront of many recent studies, particularly those examining its radiative effects, is mineral dust. The large spatial coverage of desert source regions and the fact that dust can radiatively interact with such a large part of the electromagnetic spectrum due to its range in particle size, makes it an important aerosol to study. Dust can directly scatter and absorb solar and infrared radiation which can subsequently alter the amount of radiation that would otherwise be present in the absence of dust at any level of the atmosphere like the surface. This is known as radiative forcing. At the surface dust can block incoming solar energy, however at infrared wavelengths, dust acts to partially compensate the solar losses. Evaluating the solar radiative effect of dust aerosols is relatively straightforward due in part to the relatively large signal-to-noise ratio in the measurements. At infrared wavelengths, on the other hand, the effect is rather difficult to ascertain since the measured dust signal level is on the same order as the instrumental uncertainties. Although the radiative impact of dust is much smaller in the infrared, it can still have a noticeable influence on the distribution of energy in the Earth-atmosphere system. This is mainly attributed to the strong light-absorptive properties commonly found in many earth minerals.

Trace gases and aerosols were measured in Zhangye (39.082degN, 100.276degE, 1460 m a.s. 1.), a rural site near the Gobi deserts in northwestern China during spring 2008. Primary trace gases (CO:265 ppb; SO2:3.4 ppb; NO(*y): 4.2 ppb; hereafter results given as means of hourly data) in the area were lower than in eastern China, but still indicative of marked anthropogenic emissions. Sizable aerosol mass concentration (153 micro-g/cu m) and light scattering (159/Mm at 500 nm) were largely attributable to dust emissions, and aerosol light absorption (10.3/Mm at 500 nm) was dominated by anthropogenic pollution. Distinct diurnal variations in meteorology and pollution were induced by the local valley terrain. Strong daytime northwest valley wind cleaned out pollution and was replaced by southeast mountain wind that allowed pollutants to build up overnight. In the afternoon, aerosols had single scattering albedo (SSA, 500 mn) of 0.95 and were mainly of supermicron particles, presumably dust, while at night smaller particles and SSA of 0.89-0.91 were related to Pollution. The diverse local emission sources were characterized: the CO/SO2, CO/NO(y), NO(y)/SO2 (by moles), and BC/CO (by mass) ratios for small point sources such as factories were 24.6-54.2, 25.8-35.9, 0.79-1.31, and 4.1-6.1 x 10(exp -3), respectively, compared to the corresponding inventory ratios of 43.7-71.9, 23.7-25.7, 1.84-2.79, and 3.4-4.0 x 10(exp -3) for the industrial sector in the area. The mixing between dust and pollution can be ubiquitous in this region. During a dust storm shown as an example, pollutants were observed to mix with dust, causing discernible changes in both SSA and aerosol size distribution. Further interaction between dust and pollutants during transport may modify the properties of dust particles that are critical for their large-scale impact on radiation, clouds, and global biogeochemical cycles.