This paper explores multi-task learning (MTL) for face recognition. First, we propose a multi-task convolutional neural network (CNN) for face recognition, where identity classification is the main ...task and pose, illumination, and expression (PIE) estimations are the side tasks. Second, we develop a dynamic-weighting scheme to automatically assign the loss weights to each side task, which solves the crucial problem of balancing between different tasks in MTL. Third, we propose a pose-directed multi-task CNN by grouping different poses to learn pose-specific identity features, simultaneously across all poses in a joint framework. Last but not least, we propose an energy-based weight analysis method to explore how CNN-based MTL works. We observe that the side tasks serve as regularizations to disentangle the PIE variations from the learnt identity features. Extensive experiments on the entire multi-PIE dataset demonstrate the effectiveness of the proposed approach. To the best of our knowledge, this is the first work using all data in multi-PIE for face recognition. Our approach is also applicable to in-the-wild data sets for pose-invariant face recognition and achieves comparable or better performance than state of the art on LFW, CFP, and IJB-A datasets.
•A newly improved under-load cascading failure model are established.•The method to assess the network robustness and resilience are proposed.•The threshold of key parameter to protect global energy ...and food security is obtained.•There is no positive correlation between network robustness and resilience.•We should pay more attention to the little and countries to handle the risk.
The Russian- Ukraine War has and further influence the global energy and food security. However, the detailed influence degree, key weak points and influence process is still unclear in the current. Therefore, this study established a newly improved under-load cascading failure model with consideration of overload limitation, and used it to evaluate influence of Russian- Ukraine War on the global energy and food security. This study also proposed a method to assess the network structure characteristic including robustness and resilience through model simulation under different scenarios. The main results include: The upper limitation of node load has the dominant function on the global energy and food security, while the influence of lower limitation parameter of node load has limited function. All of the networks have relative consistent recover and anti-damage ability against Russian and Ukraine War and the global panic except barley network. A key phenomenon we should concern is that the largest trade flow amounts are not occurred in the failure nodes. The failure nodes are always the countries with low economic scale and political status. The results tell us that we should further strengthen the importance of enhance production ability and energy types to resist the risk of Russian and Ukraine War. The global international organizations are also required to strengthen the function of balance the global security demand of energy and food between big countries and small countries. We should pay more attention to the little countries in the Africa and Asia to handle the risk.
In this paper we calculate the tree level three-point functi ons of Vasiliev’s higher spin gauge theory in
AdS
4
and find agreement with the correlators of the free field theory of
N
massless scalars ...in three dimensions in the
O
(
N
) singlet sector. This provides substantial evidence that Vasiliev theory is dual to the fre e field theory, thus verifying a conjecture of Klebanov and Polyakov. We also find agreement with the critical
O
(
N
) vector model, when the bulk scalar field is subject to the alternative boundary condition such that its dual operator has classical dimension 2.
Pedestrian detection is a critical problem in computer vision with significant impact on safety in urban autonomous driving. In this work, we explore how semantic segmentation can be used to boost ...pedestrian detection accuracy while having little to no impact on network efficiency. We propose a segmentation infusion network to enable joint supervision on semantic segmentation and pedestrian detection. When placed properly, the additional supervision helps guide features in shared layers to become more sophisticated and helpful for the downstream pedestrian detector. Using this approach, we find weakly annotated boxes to be sufficient for considerable performance gains. We provide an in-depth analysis to demonstrate how shared layers are shaped by the segmentation supervision. In doing so, we show that the resulting feature maps become more semantically meaningful and robust to shape and occlusion. Overall, our simultaneous detection and segmentation framework achieves a considerable gain over the state-of-the-art on the Caltech pedestrian dataset, competitive performance on KITTI, and executes 2 × faster than competitive methods.
Fish biology has been developed for more than 100 years, but some important breakthroughs have been made in the last decade. Early studies commonly concentrated on morphology, phylogenetics, ...development, growth, reproduction manipulation, and disease control. Recent studies have mostly focused on genetics, molecular biology, genomics, and genome biotechnologies, which have provided a solid foundation for enhancing aquaculture to ensure food security and improving aquatic environments to sustain ecosystem health. Here, we review research advances in five major areas: (1) biological innovations and genomic evolution of four significant fish lineages including non-teleost ray-finned fishes, northern hemisphere sticklebacks, East African cichlid fishes, and East Asian cyprinid fishes; (2) evolutionary fates and consequences of natural polyploid fishes; (3) biological consequences of fish domestication and selection; (4) development and innovation of fish breeding biotechnologies; and (5) applicable approaches and potential of fish genetic breeding biotechnologies. Moreover, five precision breeding biotechniques are examined and discussed in detail including gene editing for the introgression or removal of beneficial or detrimental alleles, use of sex-specific markers for the production of mono-sex populations, controllable primordial germ cell on-off strategy for producing sterile offspring, surrogate broodstock-based strategies to accelerate breeding, and genome incorporation and sexual reproduction regain-based approach to create synthetic polyploids. Based on these scientific and technological advances, we propose a blueprint for genetic improvement and new breed creation for aquaculture species and analyze the potential of these new breeding strategies for improving aquaculture seed industry and strengthening food security.
A
bstract
We treat RR flux backgrounds of type II string theory in the framework of closed superstring field theory based on the NSR formalism, focusing on two examples: (1) the pp-wave background ...supported by 5-form flux, and (2) AdS
3
×
S
3
×
M
4
supported by mixed 3-form fluxes. In both cases, we analyze the classical string field solution perturbatively, and compute the correction to the dispersion relation of string states to quadratic order in the RR flux. In the first example, our result is in a delicate way consistent with that obtained from lightcone quantization of the Green-Schwarz string. In the second example, we will obtain numerically the mass corrections to pulsating type IIB strings in AdS
3
×
S
3
×
M
4
. Our results, valid at finite AdS radius, agree with previously known answers in the semiclassical limit and in the BMN limit respectively.
The large pose discrepancy between two face images is one of the key challenges in face recognition. Conventional approaches for pose-invariant face recognition either perform face frontalization on, ...or learn a pose-invariant representation from, a non-frontal face image. We argue that it is more desirable to perform both tasks jointly to allow them to leverage each other. To this end, this paper proposes Disentangled Representation learning-Generative Adversarial Network (DR-GAN) with three distinct novelties. First, the encoder-decoder structure of the generator allows DR-GAN to learn a generative and discriminative representation, in addition to image synthesis. Second, this representation is explicitly disentangled from other face variations such as pose, through the pose code provided to the decoder and pose estimation in the discriminator. Third, DR-GAN can take one or multiple images as the input, and generate one unified representation along with an arbitrary number of synthetic images. Quantitative and qualitative evaluation on both controlled and in-the-wild databases demonstrate the superiority of DR-GAN over the state of the art.
Heavy metal contamination in soils has worsened with rapid economic development. The combined method with principal component analysis/absolute principal component scores) and random forest models ...successfully reveals the total sources contribution structure and the specific influence process of industrial activities on heavy metals concentration in soils of the three urban agglomerations. Through statistical analysis, Cd in the Jing-Jin-Ji Metropolitan Region, Pb, Cu, Hg, Cd and As in the Yangtze River Delta, and Hg, As, Cd, Cu and Pb in the Pearl River Delta had relatively high mean concentrations and coefficient of variation (CV), which indicates that the contamination may be caused by human activities. Through PCA/APCS analysis, industrial activities contributed more than 60% of the Cd, Hg and Pb concentrations in soils in JJJ, YRD and PRD. A random forest simulation revealed that heavy metal pollution in soils is the combined result of natural processes and human activities in the three urban agglomerations. The heavy metals concentration in JJJ is mainly caused by industrial activities through land-based emission. The industrial activities presented more significant impact on heavy metals concentration in the soil of YRD compared with other two urban agglomerations. The elevation variation controlled the pattern of heavy metal concentration through influencing the spatial clustering feature of industrial activities. The distance from the sample location to the nearest industrial enterprise is the most important factor in determining the heavy metal concentration. The number of enterprises within a 5 km radius of the sample locations makes a greater contribution to the amount of Hg pollution than other heavy metals. The results of this study could provide support for better management of soil pollution prevention practices such as specific industrial governance and layout optimization.
Linking water to research on coupled human and natural systems (CHANS) has attracted wide interest as a means of supporting human-natural sustainability. However, most current research does not focus ...on water environmental properties; instead, it is at the stage of holistic status assessment and measures adjustment from the point of view of the whole study region without revealing the dynamic interaction between human activities and natural processes. This paper establishes an integrated model that combines a System Dynamics model, a Cell Automaton model and a Multiagent Systems model and exploits the potential of the combined model to reveal regions' human-water interaction status during the process of urban evolution, identify the main pollution sources and spatial units, and provide the explicit space-time measurements needed to enhance local human-natural sustainability. The successful application of the integrated model in the case study of Changzhou City, China reveals the following. (1) As the city's development has progressed, the water environment status in some spatial units is still unsatisfactory and may even become more serious, especially in the urban areas of the Urban District and Liyang County. The concentration of Chemical Oxygen Demand (COD) in monitoring section 157 of the Urban District has increased from 36.90 mg/l to 40.84 mg/l. The main source of this increase is the increase in secondary industry. (2) With the application of the spatially explicit measures of the sewage treatment ratio improvement and new sewage plant construction, the water quality in the urban area has significantly improved and now satisfies the water quality standards. The measure of livestock manure utilization enhancement is adopted to improve the spatial units in which livestock is the main pollution source and achieve the goal of water quality improvement. The model can be used to support the sustainable status assessment of human-water interaction and to identify effective measures that can be used to realize human-water sustainability along with social-economic development.
Display omitted
•An integrated model has been established for human-water simulation.•It could reveal human-water dynamic change along with urban evolution.•It realizes pollution sources structure analysis for a specific monitoring section.•It could support for spatial explicit adjustment for human-water sustainability.
A spatially explicit approach is proposed to reveal human-water dynamic change and support human-water sustainable management.
A
bstract
We explore the analytic structure of the non-perturbative S-matrix in arguably the simplest family of massive non-integrable quantum field theories: the Ising field theory (IFT) in two ...dimensions, which may be viewed as the Ising CFT deformed by its two relevant operators, or equivalently, the scaling limit of the Ising model in a magnetic field. Our strategy is that of collider physics: we employ Hamiltonian truncation method (TFFSA) to extract the scattering phase of the lightest particles in the elastic regime, and combine it with S-matrix bootstrap methods based on unitarity and analyticity assumptions to determine the analytic continuation of the 2 → 2 S-matrix element to the complex
s
-plane. Focusing primarily on the “high temperature” regime in which the IFT interpolates between that of a weakly coupled massive fermion and the
E
8
affine Toda theory, we will numerically determine 3-particle amplitudes, follow the evolution of poles and certain resonances of the S-matrix, and exclude the possibility of unknown wide resonances up to reasonably high energies.