The modulation of diurnal cycle of precipitation over the Maritime Continent (MC) by the Madden–Julian Oscillation (MJO) during boreal winter was studied for the period of 1998–2014 using TRMM ...rainfall and MERRA reanalysis data. The RMM index by Wheeler and Hendon is used to define MJO phases. It is found that there are statistically significant differences in the diurnal cycle amplitude between different MJO phases. The amplitude of precipitation diurnal cycle is largest (smallest) during phases 2–3 (6–7), especially for the diurnal cycle over the western MC. Analysis of the local precipitation budget indicates that vertically integrated horizontal moisture convergence plays an important role in causing the amplitude difference. A further diagnosis indicates that the difference is primarily attributed to the convergence of the MJO-scale moisture by diurnal wind. The enhanced diurnal wind activity during phases 2–3 results from enhanced condensational heating on the diurnal scale, owing to the increase of surface moisture caused by MJO induced boundary layer convergence in front of convection. The increase of the background low-level moisture associated with MJO further promotes a positive circulation–convection feedback to enhance the diurnal cycle activity. Opposite processes operate at phases 6–7.
About 250 thermokarst lakes, with an average size of 5580
m
2 and a total size of 139
×
10
4
m
2, are spread between the Kunlun Mountain pass and the Fenghuo Mountain pass along the Qinghai–Tibet ...railway, where ice-rich and warm permafrost exists. Approximately 56% of the lakes are elliptical and 23% are elongated. The water depth varies between 0.4 and 3
m. Most of the lakes in the Chumarhe High Plain and other mountain regions are shallower than the thickness of winter ice (60 to 80
cm), and are frozen to the bottom during the cold seasons. However, in the Hoh Xil Hill region and Beiluhe basin, the water depth is greater than the thickness of winter ice and the relatively warm lake bottoms can cause considerable disturbance to the surrounding continuous permafrost. The lakes in the Chumarhe High Plain are saltwater or brine lakes; and in other regions, the lakes contain freshwater or brackish water. Ages of formation, enlargement rates, water and lake-bottom temperatures, the configuration of permafrost, and active-layer thickness were measured at a typical thermokarst lake in the Beiluhe basin between 2007 and 2009. The lake formed about 890
years ago and is 150
m long, 100
m wide, with a water depth of 2.5
m now. The size of the lake is growing at a rate of about 1
m/a at some shores. The lakeshores are underlain by permafrost, which terminates almost vertically at their edge. The mean annual temperature measured at lake bottom in the centre was 5.45
°C, while the temperature of permafrost at 15
m depth in the lakeshore varied from −
0.84
°C near the edge of the lake to −
1.26
°C at 80
m from the edge. The corresponding active layer depths varied from 2.45 to 1.80
m. The configuration of the talik indicates that no permafrost is found beneath the centre of the lake.
► The characteristics of the thermokarst lakes along the Qinghai-Tibet Engineering Corridor have been researched. ► The existence of the lakes is closely related to the permafrost ice-content and temperature. ► The enlargement of size, the age, the mean annual temperatures at the lake surface and lake bottom, and the process of the ice freezing and melting of a lake in Beiluhe basin have also measured. ► The thermokarst lake would bring great impact on the permafrost configuration beneath and around it.
•The proposed method introduces the deterministic policy into the generative adversarial imitation learning method, so that the robot can quickly imitate the demonstration policy without reward ...engineering.•Drawing on the idea of learning from demonstration (LfD), the method inputs the demonstration data both into the generator and the discriminator in order to improve the stability of training process by leveraging from the demonstration data rather than simply from the exploration data.
Display omitted
This paper proposes a deterministic generative adversarial imitation learning method which allows the robot to implement the motion planning task rapidly by learning from the demonstration data without reward function. In our method, the deep deterministic policy gradient method is used as the generator for learning the action policy on the basis of discriminator, and the demonstration data is input into the generator to ensure its stability. Three experiments on the push and pick-and-place tasks are conducted in the gym robotic environment. Results show that the learning speed of our method is much faster than the stochastic generative adversarial imitation learning method, and it can effectively learn from the demonstration data in different states of the task with higher learning stability. The proposed method can complete the motion planning task without environmental reward quickly and improve the stability of the training process.
Despite current advancement in the field of biomedical image processing, propelled by the deep learning revolution, multimodal image registration, due to its several challenges, is still often ...performed manually by specialists. The recent success of image-to-image (I2I) translation in computer vision applications and its growing use in biomedical areas provide a tempting possibility of transforming the multimodal registration problem into a, potentially easier, monomodal one. We conduct an empirical study of the applicability of modern I2I translation methods for the task of rigid registration of multimodal biomedical and medical 2D and 3D images. We compare the performance of four Generative Adversarial Network (GAN)-based I2I translation methods and one contrastive representation learning method, subsequently combined with two representative monomodal registration methods, to judge the effectiveness of modality translation for multimodal image registration. We evaluate these method combinations on four publicly available multimodal (2D and 3D) datasets and compare with the performance of registration achieved by several well-known approaches acting directly on multimodal image data. Our results suggest that, although I2I translation may be helpful when the modalities to register are clearly correlated, registration of modalities which express distinctly different properties of the sample are not well handled by the I2I translation approach. The evaluated representation learning method, which aims to find abstract image-like representations of the information shared between the modalities, manages better, and so does the Mutual Information maximisation approach, acting directly on the original multimodal images. We share our complete experimental setup as open-source (https://github.com/MIDA-group/MultiRegEval), including method implementations, evaluation code, and all datasets, for further reproducing and benchmarking.
Despite current advancement in the field of biomedical image processing, propelled by the deep learning revolution, multimodal image registration, due to its several challenges, is still often ...performed manually by specialists. The recent success of image-to-image (I2I) translation in computer vision applications and its growing use in biomedical areas provide a tempting possibility of transforming the multimodal registration problem into a, potentially easier, monomodal one. We conduct an empirical study of the applicability of modern I2I translation methods for the task of rigid registration of multimodal biomedical and medical 2D and 3D images. We compare the performance of four Generative Adversarial Network (GAN)-based I2I translation methods and one contrastive representation learning method, subsequently combined with two representative monomodal registration methods, to judge the effectiveness of modality translation for multimodal image registration. We evaluate these method combinations on four publicly available multimodal (2D and 3D) datasets and compare with the performance of registration achieved by several well-known approaches acting directly on multimodal image data. Our results suggest that, although I2I translation may be helpful when the modalities to register are clearly correlated, registration of modalities which express distinctly different properties of the sample are not well handled by the I2I translation approach. The evaluated representation learning method, which aims to find abstract image-like representations of the information shared between the modalities, manages better, and so does the Mutual Information maximisation approach, acting directly on the original multimodal images. We share our complete experimental setup as open-source (https://github.com/MIDA-group/MultiRegEval), including method implementations, evaluation code, and all datasets, for further reproducing and benchmarking.
Sign language recognition aims to recognize meaningful movements of hand gestures and is a significant solution in intelligent communication between the deaf community and hearing societies. However, ...until now, the current dynamic sign language recognition methods still have some drawbacks with difficulties of recognizing complex hand gestures, low recognition accuracy for most dynamic sign language recognition, and potential problems in larger video sequence data training. In order to solve these issues, this paper presents a multimodal dynamic sign language recognition method based on a deep 3-dimensional residual ConvNet and bi-directional LSTM networks, which is named as BLSTM-3D residual network (B3D ResNet). This method consists of three main parts. First, the hand object is localized in the video frames in order to reduce the time and space complexity of network calculation. Then, the B3D ResNet automatically extracts the spatiotemporal features from the video sequences and establishes an intermediate score corresponding to each action in the video sequence after feature analysis. Finally, by classifying the video sequences, the dynamic sign language is accurately identified. The experiment is conducted on test datasets, including DEVISIGN_D dataset and SLR_Dataset. The results show that the proposed method can obtain state-of-the-art recognition accuracy (89.8% on the DEVISIGN_D dataset and 86.9% on SLR_Dataset). In addition, the B3D ResNet can effectively recognize complex hand gestures through larger video sequence data, and obtain high recognition accuracy for 500 vocabularies from Chinese hand sign language.
The FKM-g-S
i
R macromolecular compatibilizer was produced by using a high-temperature thermomechanical grafting process to graft long-chain vinyl silicone oils (S
i
R, terminal vinyl dimethicone) ...with varying vinyl contents (0.09%,0.21%, and 0.63%) onto fluoroelastomer (FKM). The impact of initiator dosage, silicone oil dosage and vinyl content on the mechanical characteristics of FKM/MVQ (Methyl vinyl silicone rubber) blends was probed using a three-level orthogonal design experiment with a three-factor range analysis. Optimal conditions for the preparation of the compatibilizer were obtained accordingly. The findings indicated that the vulcanization properties, mechanical properties, and resistance to oil of the FKM/MVQ blends that were rendered compatible were optimized when the quantity of initiator Dicumyl Peroxide (DCP) was 0.25 parts, the volume of silicone oil was 9 parts, and the vinyl content was 0.63%. The composites exhibited the following specifications: the maximum torque difference can reach up to 12.56 dN.m, while the tensile strength and elongation at break can reach up to 8.39 MPa and 348.04%, respectively. Additionally, the rate of change in oil volume and mass can be reduced by 25.2% and 14.37%, correspondingly. The blended rubber's structure and performance were characterized using several methods, including Fourier transform infrared spectroscopy (FT-IR), Rubber processing analyzer (RPA), Scanning electron microscopy (SEM), Transmission electron microscopy (TEM) and Dynamic mechanical thermal analysis (DMTA). The results indicated that the compatibilizer significantly improved the compatibility of the FKM/MVQ blend.
Physical activity intervention and watching natural environment videos have been proven to improve young children's attention levels. However, evidence comparing the improvement effects of different ...combinations of the two activities has rarely been reported. By comparing the differences in the improvement effects of four combinations of physical activities and watching natural environment videos on young children's attention levels, this study can enrich the evidence in this research field and also provide a reference for arranging effective intervention methods for children's attention recovery between classes.
A total of 152 children aged 4 to 6 years were recruited and randomly divided into four intervention groups: (1) physical activity intervention first and thereafter watching a natural environment video group (activity + video group), (2) watching a natural environment video first and thereafter the physical activity intervention group (video + activity group), (3) physical activity-based group, and (4) natural environment video-based group. Physical activity involved 4 min of moderate-intensity basic physical fitness combination training. The subjects wore the Pico Neo pioneer version of the VR glasses all-in-one machine to watch a natural environment video. Thereafter, population sociological variables and daily physical activity levels were investigated. Auditory and visual sustained attention tests were performed before and after intervention in each group.
The auditory attention post-test scores of the four groups showed an improvement trend compared with the pretest scores. In particular, the activity + video group (F = 10.828; ɳp2 = 0.226; p = 0.002) and natural environment video-based group (F = 9.452; ɳp2 = 0.203; p = 0.004) have the best improvement effect. For visual attention, only the activity + video group showed a significant improvement trend (F = 4.287; ɳp2 = 0.104; p = 0.045), while the other three groups showed a downward trend in scores.
Among the different intervention combinations, the physical activity intervention first and watching natural environment videos thereafter group has the best effect on improving children's attention during recess. Physical activity interventions at the end of recess adversely affect young children's visual attention levels at the beginning of the class. Therefore, this study recommends that children should not engage in physical activity interventions in the second half of the class break. Lastly, the current research recommends presenting the content of physical activity interventions first and further improving their attention thereafter by watching natural environment videos.
As a widely used core component of hydropower generation, the stable and safe operation of the Francis turbine plays a very important role in engineering applications. Therefore, the effects of ...different rotor blade numbers on the hydraulic performance and flow control of the draft tube of the Francis turbine are of great research value. This paper focuses on the rotor blade number 13, 14, 15, 16, and 17 in the prohibited and stabilized operating zones, each of which is taken as a working condition point. (Condition αref -HM and Condition 2αref -HM). Computational Fluid Dynamics (CFD) numerical simulations were performed and analyzed. In terms of hydraulic performance, the increase in the number of blades increases the efficiency and power of the turbine under Condition αref -HM by 2.26% and 0.0021 MW respectively. The efficiency and power of the turbine under Condition αref -HM is also elevated, with the efficiency reaching a maximum of 91.77% at blade number 15, and the power of the same turbine reaching a maximum of 0.182 MW at runner blade number 13. From the 3D flow analysis of the turbine, the increase in the number of blades does not significantly change the flow state of the turbine’s draft tube. By defining the turbulence energy E, the fast Fourier transform of the turbulence energy signals on the two monitoring surfaces of the draft tube is used to obtain its main frequency, and the amplitude and phase at the typical main frequency are visualized, through which we can analyze the vortex band motion state on the two monitoring surfaces of the draft tube, and speculate the change of pressure pulsation of the draft tube. Such an analysis method has a guiding significance for us to improve the performance and stability of the turbine, which can be achieved by increasing or decreasing the number of blades.
The goal of reinforcement learning is to enable an agent to learn by using rewards. However, some robotic tasks naturally specify with sparse rewards, and manually shaping reward functions is a ...difficult project. In this article, we propose a general and model-free approach for reinforcement learning to learn robotic tasks with sparse rewards. First, a variant of Hindsight Experience Replay, Curious and Aggressive Hindsight Experience Replay, is proposed to improve the sample efficiency of reinforcement learning methods and avoid the need for complicated reward engineering. Second, based on Twin Delayed Deep Deterministic policy gradient algorithm, demonstrations are leveraged to overcome the exploration problem and speed up the policy training process. Finally, the action loss is added into the loss function in order to minimize the vibration of output action while maximizing the value of the action. The experiments on simulated robotic tasks are performed with different hyperparameters to verify the effectiveness of our method. Results show that our method can effectively solve the sparse reward problem and obtain a high learning speed.
PDF
