Drosophila (Sophophora) subpulchrella Takamori and Watabe, sp. nov., of the D. suzukii subgroup in the D. melanogaster species group, is described from Japan and southern China, and compared with its sibling species, D. pulchrella Tan et al. distributed in the Yun-Gui Highland, south-western China. The results of cross-experiments show a complete pre-mating isolation between D. subpulchrella and D. pulchrella.
Deep Neural Networks (DNNs) have revolutionized a wide range of industries, from healthcare and finance to automotive, by offering unparalleled capabilities in data analysis and decisionmaking. Despite their transforming impact, DNNs face two critical challenges: the vulnerability to adversarial attacks and the increasing computational costs associated with more complex and larger models. In this paper, we introduce an effective method designed to simultaneously enhance adversarial robustness and execution efficiency. Unlike prior studies that enhance robustness via uniformly injecting noise, we introduce a non-uniform noise injection algorithm, strategically applied at each DNN layer to disrupt adversarial perturbations introduced in attacks. By employing approximation techniques, our approach identifies and protects essential neurons while strategically introducing noise into non-essential neurons. Our experimental results demonstrate that our method successfully enhances both robustness and efficiency across several attack scenarios, model architectures, and datasets.
Despite their important contribution to the economic domain, active heat-releasing industrial plants have significant implications for human health and climate change. However, a spatially detailed dataset of various heat-releasing industrial sectors and large-scale characterization of heat emissions from industrial sources have not been reported yet. In this study, a dataset of heat-releasing industries was established using a national detection map of thermal anomalies produced by a novel and more accurate method employing daily nighttime visible infrared imaging radiometer suite thermal infrared images corresponding to 1 year. Subsequently, we quantified the dimensional features of heat radiation fluxes of China's industrial plants. A total of 12 114 industrial objects were structured in a two-level hierarchical dataset of heat-releasing industries, representing a magnitude of at least 1 order higher than the number enumerated in the state-of-the-art inventory of industrial heat sources across China. The satellite observations helped more completely characterize industrial heat plumes, which represent the industrial heat radiation fluxes with higher levels of densities that prevail in the central-eastern part of China having spatial clustering islands. Our results could be used to inform policy and environmental management in relation to meaningful dynamic industrial supervision, targeting extreme polluters and differentiated emission mitigation measurements.
Eastern and Western Asia were important centers for the domestication of plants and animals and they developed different agricultural practices and systems. The timing, routeway and mechanisms of the exchanges between the two centers have long been important scientific issues. The development of a mixed pastoral system (e.g., with the rearing of sheep, goats and cattle) and millet cultivation in the steppe region of northern China was the result of the link between the two cultures. However, little detailed information is available about the precise timing and mechanisms involved in this mixture of pastoralism and millet cultivation. To try to address the issue, we analyzed the pollen, fungal spores and phytolith contents of soil samples from the Bronze Age Zhukaigou site in the steppe area of North China, which was combined with AMS 14C dating of charcoal, millet and animal bones. A mixed pastoralism and millet agricultural system appeared at the site between 4,000 and 3,700 cal yr BP, and the intensity of animal husbandry increased in the later stage of occupation. Published data indicate that domestic sheep/goats appeared across a wide area of the steppe region of northern China after ∼4,000 cal yr BP. A comparison of records of sheep/goat rearing and paleoclimatic records from monsoon area in China leads us to conclude that the mixture of pastoralism and millet cultivation was promoted by the occurrence of drought events during 4,200–4,000 cal yr BP. Moreover, we suggest that mixed rainfed agriculture and animal husbandry increased the adaptability and resilience of the inhabitants of the region which enabled them to occupy the relatively arid environment of the monsoon marginal area of northern China.
In realistic environments, speech is usually interfered by various noise and reverberation, which dramatically degrades the performance of automatic speech recognition (ASR) systems. To alleviate this issue, the commonest way is to use a well-designed speech enhancement approach as the front-end of ASR. However, more complex pipelines, more computations and even higher hardware costs (microphone array) are additionally consumed for this kind of methods. In addition, speech enhancement would result in speech distortions and mismatches to training. In this paper, we propose an adversarial training method to directly boost noise robustness of acoustic model. Specifically, a jointly compositional scheme of generative adversarial net (GAN) and neural network-based acoustic model (AM) is used in the training phase. GAN is used to generate clean feature representations from noisy features by the guidance of a discriminator that tries to distinguish between the true clean signals and generated signals. The joint optimization of generator, discriminator and AM concentrates the strengths of both GAN and AM for speech recognition. Systematic experiments on CHiME-4 show that the proposed method significantly improves the noise robustness of AM and achieves the average relative error rate reduction of 23.38% and 11.54% on the development and test set, respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.