Search

Neural network models for hyperspectral images classification are complex and therefore difficult to deploy directly onto mobile platforms. Neural network model compression methods can effectively optimize the storage space and inference time of the model while maintaining the accuracy. Although automated pruning methods can avoid designing pruning rules, they face the problem of search efficiency when optimizing complex networks. In this paper, a network collaborative pruning method is proposed for hyperspectral image classification based on evolutionary multi-task optimization. The proposed method allows classification networks to perform the model pruning task on multiple hyperspectral images simultaneously. Knowledge (the important local sparse structure of the network) is automatically searched and updated by using knowledge transfer between different tasks. The self-adaptive knowledge transfer strategy based on historical information and dormancy mechanism is designed to avoid possible negative transfer and unnecessary consumption of computing resources. The pruned networks can achieve high classification accuracy on hyperspectral data with limited labeled samples. Experiments on multiple hyperspectral images show that the proposed method can effectively realize the compression of the network model and the classification of hyperspectral images.

It is widely known that an optimal nucleotide sequence context immediately upstream of the AUG start codon greatly improves the efficiency of translation initiation of mRNA in mammalian and plant somatic cells, which in turn increases protein levels. However, it is still unclear whether a similar regulatory mechanism is also present in highly differentiated cells. Here, we surveyed this issue in Arabidopsis thaliana sperm cells and found that the sequence context-mediated regulation of translation initiation in sperm cells is generally similar to that in somatic cells. A simple motif of four adenine nucleotides at positions - 1 to - 4 greatly improved the efficiency of translation initiation, and when the motif was present there, translation was even initiated at some non-AUG codons in sperm cells. However, unlike that in mammalian cells, a mainly effective nucleotide site to regulate the efficiency of translation initiation was not present at positions - 1 to - 4 in sperm cells. Meanwhile, different from somatic cells, sperm cells did not use eukaryotic translation initiation factor 1 to regulate the efficiency in a poor context consisting of the lowest frequency nucleotides. All these results contribute to our understanding of the cytoplasmic event of translation initiation in highly differentiated sperm cells.

Water and fertilizer play an important role in crop growth in dryland areas. It is a necessity to improve the water use efficiency (WUE) of the crop once the water resource is limited. In northern China, where there is a wide shortage of water resources, it is therefore necessary to investigate how fertilization affects the WUE of spring maize and to quantify the effects. A total of 33 published peer-reviewed papers were collected, and a meta-analysis and random forest model analysis were performed with 364 WUE comparisons, aiming to explore the effects of fertilization on the WUE of spring maize and to clarify the optimal conditions for WUE under fertilizer management. The results showed that fertilization significantly increased the WUE of spring maize by 56.72% (P < 0.01) when compared with non-fertilization. The WUE effect under the organic–inorganic fertilizer combination (MNPK) was approximately twice as high as that under inorganic fertilizer (NPK) or organic fertilizer (M). The greatest increase in WUE occurred at 0–100 kg ha−1 of nitrogen application (NA). Under environmental conditions including 7 ≤ mean annual temperature in the test year (T) ≤ 10 °C, 400 ≤ mean annual precipitation in the test year (P) ≤ 600 mm, and mean altitude (A) > 1500 m, and soil conditions including 10 ≤ soil organic matter content (SOM) ≤ 14 g kg−1 and available phosphorus (AP) < 5 mg kg−1, the fertilization optimally enhanced the WUE of spring maize when the agronomic measures of ridge–furrow planting (RFP) and mulching film (MF) were used. The random forest model analysis indicated that the influence factors (i.e., fertilizer regimes, environmental factors, soil factors, and agronomic measures) caused 65.62% of the variation in spring maize WUE effects, while in all influence factors, fertilizer types related to fertilizer regimes caused the most variation. The initial available potassium (AK) and available nitrogen (AN) of the soil were negatively correlated to the WUE effect, indicating that fertilization imposed a better effect on the WUE of spring maize when the soil was infertile. Fertilization significantly increased the WUE of spring maize, and organic and inorganic fertilizer application provided an effective measure for the sustainable development of spring maize in northern China. After clarifying the required conditions for fertilization increasing WUE, high-efficiency water use may be achieved.

We propose a CNT-based concentric twin tube (CTT) as nanochannels for both water purification and ion separation at the nanoscale. In the model, a source reservoir dealing with the solution connects three containers via the CTT that has three subchannels for mass transfer. Before entering the three subchannels, the solution in the separating zone will form three layers (the aqua cations, water, and the aqua anions, respectively) by applying a charged capacitor with the two electrodes parallel to the flow direction of the solution. Under an electric field with moderate intensity, the three subchannels in the CTT have stable configurations for mass transfer. Since the water and the two types of aqua ions are collected by three different containers, the present model can realize both ion separation and water purification. The mass transfer in the subchannels will be sped up by an external pressure exerted on the solution in the source reservoir. The physical properties of the model, e.g., water purification speed, are analyzed with respect to the effects of the electric field, the size of CTT, and the concentration of solute, such as NaCl.

Due to its pleasant rose-like scent, 2-phenylethanol (2-PE) has been widely used in the fields of cosmetics and food. Microbial production of 2-PE offers a natural and sustainable production process. However, the current bioprocesses for de novo production of 2-PE suffer from low titer, yield, and productivity. In this work, a multilevel metabolic engineering strategy was employed for the high-level production of 2-PE. Firstly, the native alcohol dehydrogenase YugJ was identified and characterized for 2-PE production via genome mining and gene function analysis. Subsequently, the redirection of carbon flux into 2-PE biosynthesis by combining optimization of Ehrlich pathway, central metabolic pathway, and phenylpyruvate pathway enabled the production of 2-PE to a titer of 1.81 g/L. Specifically, AroK and AroD were identified as the rate-limiting enzymes of 2-PE production through transcription and metabolite analyses, and overexpression of aroK and aroD efficiently boosted 2-PE synthesis. The precursor competing pathways were blocked by eliminating byproduct formation pathways and modulating the glucose transport system. Under the optimal condition, the engineered strain PE23 produced 6.24 g/L of 2-PE with a yield and productivity of 0.14 g/g glucose and 0.13 g/L/h, respectively, using a complex medium in shake flasks. This work achieves the highest titer, yield, and productivity of 2-PE from glucose via the phenylpyruvate pathway. This study provides a promising platform that might be widely useful for improving the production of aromatic-derived chemicals.

Traditional band selection methods only analyze one data set at a time, and start searching band subsets from the zero ground state of knowledge, which can not effectively mine spectral information to guide band selection. However, for hyperspectral images obtained by the same sensor, the spectral information has similar physical meaning (radiance or reflectivity). Collaborative analysis technology can analyze multiple hyperspectral datasets to explore the inherent spectral features shared among them. In this paper, a multiple data sets collaborative analysis framework for hyperspectral band selection is proposed to realize spectral information communication, thereby guiding and promoting band selection of each data set. Different band selection tasks are established pertinently, then the evolutionary multitasking band selection method is designed to facilitate the knowledge sharing of different band selection tasks. More importantly, the interaction mechanism among different data sets is adjusted dynamically, thereby improving the cooperation ability of the collaborative analysis framework. Besides, a predominant gene reservation crossover and a deduplication mutation are designed for retaining the promising bands and avoiding the selection of repeat bands. Experiments indicate that the proposed collaborative analysis method works more efficiently than the comparison methods and successfully enhances accuracy and convergence compared to single data set analysis.

BiOI nanosheets could catalyze the photothermal catalytic CO2 oxidative dehydrogenation of propane to propylene.

With abundant temporal, spectral, and spatial information, multispectral images are proficient for acquiring a superior comprehension of the Earth's condition and its changes, which enables the achievement of multiple change detection (CD) tasks. However, high temporal, spatial, and spectral information of data brings obstacles to perform multiple change analysis due to the lack of effective feature extraction operation. In addition, the traditional multiple CD methods rely too much on manual participation. Here, a generative representation learning network (GRN) and a cyclic clustering technique are combined into a unified model, which is driven to learn spatial-temporal-spectral features for unsupervised multiple CD. GRN aims to efficiently extract and merge robust difference information with a recurrent learning mechanism for self-adaptive classification refinement, in which different types of changes can be identified and highlighted. Furthermore, a cyclic training strategy is designed to refine the clustering-friendly features, in which similar change types are gradually merged into the same classes. Meanwhile, the number of change types will be optimized through a self-adaptive way and eventually converge to its stable state, which is close to the real distribution. Experimental results on real multispectral datasets demonstrate the effectiveness and superiority of the proposed model on multiple CD.

Introduction Nearly one in seven women worldwide suffers from chronic pelvic pain syndrome (CPPS) each year. Often, CPPS necessitates a combination of treatments. Studies have shown the good therapeutic effects of repetitive transcranial magnetic stimulation (rTMS) upon CPPS. We wish to undertake a randomized controlled trial (RCT) to observe the effect of high-frequency rTMS at different intensities upon CPPS. Methods and analyses In this prospective, double-blinded RCT, 63 female CPPS participants will be recruited and randomized (1:1:1) to high-intensity rTMS, low-intensity rTMS, or sham rTMS. The control group will receive a 10-day course of conventional pelvic floor (PF) rehabilitation (neuromuscular stimulation, magnetic therapy, or light therapy of the PF). On the basis of conventional treatment, participants in the high-intensity rTMS group will receive pulses of 10 Hz with a resting motor threshold (RMT) of 110% for a total of 15,000 pulses. Participants in the low-intensity rTMS group will receive pulses of 10 Hz with an RMT of 80% with 15,000 pulses. The sham rTMS group will be subjected to sham stimulation with the same sound as produced by the real magnetic stimulation coil. The primary outcome will be determined using a visual analog scale, the Genitourinary Pain Index, Zung Self-Rating Anxiety Scale, and Zung Self-Rating Depression Scale. The secondary outcome will be determined by electromyography of the surface of PF muscles at baseline and after treatment completion. Ethics and dissemination This study is approved by the Ethics Committee of Bao’an People’s Hospital, Shenzhen, Guangdong Province (approval number: BYL20211203). The results will be submitted for publication in peer-reviewed journals and disseminated at scientific conferences (Protocol version 1.0-20220709). Trial registration Chictr.org.cn, ID: ChiCTR2200055615. Registered on 14 January 2022, http://www.chictr.org.cn/showproj.aspx?proj=146720. Protocol version 1.0-20220709.

Graph neural networks (GNNs) have shown the power in representation learning over graph-structured user-item interaction data for collaborative filtering (CF) task. However, with their inherently recursive message propagation among neighboring nodes, existing GNN-based CF models may generate indistinguishable and inaccurate user (item) representations due to the over-smoothing and noise effect with low-pass Laplacian smoothing operators. In addition, the recursive information propagation with the stacked aggregators in the entire graph structures may result in poor scalability in practical applications. Motivated by these limitations, we propose a simple and effective collaborative filtering model (SimRec) that marries the power of knowledge distillation and contrastive learning. In SimRec, adaptive transferring knowledge is enabled between the teacher GNN model and a lightweight student network, to not only preserve the global collaborative signals, but also address the over-smoothing issue with representation recalibration. Empirical results on public datasets show that SimRec archives better efficiency while maintaining superior recommendation performance compared with various strong baselines. Our implementations are publicly available at: https://github.com/HKUDS/SimRec., Comment: Accepted by ACM WWW 2023

Allergic rhinitis (AR) is one of the popular childhood diseases, bringing physical and metal burdens to the children and their families. The study was performed to detect common allergens eliciting AR in children, to investigate the prevalence of allergens in different age and gender cohorts, and to provide a reliable basis for clinical prevention and treatment of AR during childhood. We measured serum-specific IgE and performed inhalant and ingestion allergen examinations in 2,316 children with AR, in collaboration with BioSciTec GmbH. The prevalence of different allergens was determined according to gender, age, severity, and season. Among the 2,316 AR cases, the top five inhalant allergens were Dermatophagoides pteronyssinus (1,674 cases, 72.3%), Dermatophagoides farinae (1,520 cases, 65.6%), Blomia tropicalis (1,477 cases, 63.8%), Cockroach (602 cases, 26.0%), and Dog hair (602 cases, 26.0%). The top five ingestive allergens were Milk (1,111 cases, 48.0%), Egg white (543 cases, 23.4%), Shrimp/Crab (425 cases, 18.4%), Beef/Mutton (422 cases, 18.2%), and Egg yold (329 cases, 14.2%). AR severity analyses showed that 50.9% (1,180 cases) of D. pteronyssinus allergies were above level three, 47.9% (1,109 cases) of D. farinae allergies were above level three, only 23.3% (539 cases) of B. tropicalis allergies were level three, and B. tropicalis allergies were mainly of level 2. Other AR-inducing allergens mainly produced level one or two reactions. Regarding ingestion allergens, 7.9% (183 cases) of milk allergies and 4.7% (108 cases) of Shrimp/Crab allergies were above level three, and other allergens induced AR mainly of level one or two. The study investigated the major allergens eliciting AR in children from Guangdong, China, assessed the prevalence and severity among cohorts regarding age, gender, and season, and produced essential information on childhood AR, laying important references for AR prevention and treatment in the future.

Graphyne nanotube (GNT), as a promising one-dimensional carbon material, attracts extensive attention in recent years. However, the synthesis of GNT is still challenging even in the laboratory. This study reveals the feasibility of fabricating a GNT by self-assembling a monolayer graphyne (GY) ribbon on a carbon nanotube (CNT) via theoretical and numerical analysis. Triggered by the van der Waals force from the CNT, a GY ribbon near the tube first winds upon the tube and then conditionally self-assembles to form a GNT. The self-assembly process and result are heavily influenced by the ambient temperature, which indicates the thermal vibration of the nanosystem. Molecular dynamic simulation results address the temperature range conducive to successful self-assembly. Different types of GNTs, e.g. α-, β-, and γ-GNTs with specified chirality (armchair, zigzag, and chiral), length, and radius, can be obtained via self-assembly by controlling the geometry of the GY ribbons and temperature. The present theoretical understanding is helpful for fabricating GNTs with predefined morphology.

Iturin A is a biosurfactant with various applications, and its low synthesis capability limits its production and application development. Fatty acids play a critical role in cellular metabolism and target product syntheses, and the relationship between fatty acid supplies and iturin A synthesis is unclear. In this study, we attempted to increase iturin A production via strengthening fatty acid synthesis pathways in Bacillus amyloliquefaciens. First, acetyl-CoA carboxylase AccAD and ACP S-malonyltransferase fabD were overexpressed via promoter replacement, and iturin A yield was increased to 1.36 g/L by 2.78-fold in the resultant strain HZ-ADF1. Then, soluble acyl-ACP thioesterase derived from Escherichia coli showed the best performance for iturin A synthesis, as compared to those derived from B. amyloliquefaciens and Corynebacterium glutamicum, the introduction of which in HZ-ADF1 further led to a 57.35% increase of iturin A yield, reaching 2.14 g/L. Finally, long-chain fatty acid-CoA ligase LcfA was overexpressed in HZ-ADFT to attain the final strain HZ-ADFTL2, and iturin A yield reached 2.96 g/L, increasing by 6.59-fold, and the contents of fatty acids were enhanced significantly in HZ-ADFTL2, as compared to the original strain HZ-12. Taken together, our results implied that strengthening fatty acid supplies was an efficient approach for iturin A production, and this research provided a promising strain for industrial production of iturin A.

Ratiometric electrochemiluminescence (ECL) sensors can efficiently remove environmental interference to attain precise detection. Nonetheless, two eligible luminophores or coreactants were usually needed, increasing the complexity and restricting their practical application. In this study, a single luminophore of luminol with a single coreactant of H

Gene function studies benefit from the availability of mutants. In plants, Agrobacterium-mediated genetic transformation is widely used to create mutants. These mutants, also called transformants, contain one or several transfer-DNA (T-DNA) copies in the host genome. Quantifying the copy number of T-DNA in transformants is beneficial to assess the number of mutated genes. Here, we developed a competitive polymerase chain reaction (PCR)-based method to detect a single copy of a T-DNA insertion in transformants. The competitor line BHK- -1 that contains a single copy of competitor BHK- (BHK, Basta, Hygromycin, Kanamycin-resistant genes) was crossed with test transformants and the genomic DNA of F1 plants was subjected to competitive PCR. By analyzing the gray ratio between two PCR products, we were able to determine whether or not the test transformants contained a single copy of T-DNA insertion. We also generated the control lines BHK±1:1 and BHK±2:1 , which contain the target (BHK+ ) and competitor (BHK- ) in a ratio of 1:1 and 2:1, respectively. The ratios of their PCR products are useful references for quantitative analysis. Overall, this method is reliable and simple in experimental manipulations and can be used as a substitute for Southern-blot analysis to identify a single copy of T-DNA insertion in transformants.

In the last few years, neural networks were introduced to change detection for a better understanding of remote sensing images. However, the designs of these neural networks were time-consuming processes of trial and error, which failed to account for their validity. Thus, a simple and efficient change detection method based on network architecture search in terms of the evolutionary algorithm is proposed to deal with SAR images change detection problems. In the proposed method, an efficient gene encoding is applied to represent the unpredictable optimal depth and the number of neurons in each hidden layer. Besides, a combinatorial evaluation strategy and a self-adaptive network solution selection are designed for effective and reasonable network architectures. What is more, a hidden layer random alignment crossover operator and a drawing lots mutation operator are designed for the enhancement of diversity of network architectures. Experimental results on a few SAR image data sets demonstrate that the proposed method can generate appropriate networks to solving SAR images change detection.

The engineering practice shows that the application of grouting technology to treat underground engineering has strong applicability and is one of the most commonly used technical means at present. Based on the underground engineering, this paper introduces the research achievements of grouting materials in recent years, including cement-based grouting materials, mixed grouting materials, anti-scouring grouting materials, and ultra-fine cement grouting materials. Current demand of grouting materials in underground engineering, there exists large dosage of cement, high content, high cost, serious environmental pollution problems, such as looking for alternatives or mixed with other raw materials for preparation of cementation material become the development trend, compared with the cement grouting material, chemical grouting material with higher performance, but in smaller projects within the scope of application. How to reduce the production cost of chemical grouting materials, simplify the production process, overcome the existing toxicity, reduce environmental pollution and improve the durability of solidified body has become the bottleneck of its popularization and application. Some achievements have been made in the modification of cement or chemical materials by nanometer components, but there is still a long way to go before the large-scale application of grouting engineering.

Cervical cancer is the most common cancer among women worldwide. The diagnosis and classification of cancer are extremely important, as it influences the optimal treatment and length of survival. The objective was to develop and validate a diagnosis system based on convolutional neural networks (CNN) that identifies cervical malignancies and provides diagnostic interpretability. A total of 8496 labeled histology images were extracted from 229 cervical specimens (cervical squamous cell carcinoma, SCC, n = 37; cervical adenocarcinoma, AC, n = 8; nonmalignant cervical tissues, n = 184). AlexNet, VGG-19, Xception, and ResNet-50 with five-fold cross-validation were constructed to distinguish cervical cancer images from nonmalignant images. The performance of CNNs was quantified in terms of accuracy, precision, recall, and the area under the receiver operating curve (AUC). Six pathologists were recruited to make a comparison with the performance of CNNs. Guided Backpropagation and Gradient-weighted Class Activation Mapping (Grad-CAM) were deployed to highlight the area of high malignant probability. The Xception model had excellent performance in identifying cervical SCC and AC in test sets. For cervical SCC, AUC was 0.98 (internal validation) and 0.974 (external validation). For cervical AC, AUC was 0.966 (internal validation) and 0.958 (external validation). The performance of CNNs falls between experienced and inexperienced pathologists. Grad-CAM and Guided Gard-CAM ensured diagnoses interpretability by highlighting morphological features of malignant changes. CNN is efficient for histological image classification tasks of distinguishing cervical malignancies from benign tissues and could highlight the specific areas of concern. All these findings suggest that CNNs could serve as a diagnostic tool to aid pathologic diagnosis.

The path of crack propagation in a graphene sheet is significant for graphene patterning via the tearing approach. In this study, we evaluate the fracture properties of pre-cracked graphene during the tearing process, with consideration of the effects of the aspect ratio, loading speed, loading direction, and ambient temperatures on the crack propagation in the monolayer sheet. Some remarkable conclusions are drawn based on the molecular dynamic simulation results, i.e., a higher loading speed may result in a complicated path of crack propagation, and the propagation of an armchair crack may be accompanied by sp carbon links at high temperatures. The reason for this is that the stronger thermal vibration reduces the load stress difference near the crack tip and, therefore, the crack tip can pass through the sp link. A crack propagates more easily along the zigzag direction than along the armchair direction. The out-of-plane tearing is more suitable than the in-plane tearing for graphene patterning. The path of crack propagation can be adjusted by changing the loading direction, e.g., a rectangular graphene ribbon can be produced by oblique tearing. This new understanding will benefit the application of graphene patterning via the tearing approach.

Neurodegenerative disorders, including spinal cord injury (SCI), result in oxidative stress-induced cell damage. Morroniside (MR), a major active ingredient of the Chinese herb Shan Zhu Yu, has been shown to ameliorate oxidative stress and inflammatory response. Our previous study also confirmed that morroniside protects SK-N-SH cell line (human neuroblastoma cells) against oxidative impairment. However, it remains unclear whether MR also plays a protective role for oligodendrocytes that are damaged following SCI. The present study investigated the protective effects of MR against hydrogen peroxide (H2O2)-induced cell death in OLN-93 cells. MR protected OLN-93 cells from H2O2-induced injury, attenuated H2O2-induced increase in reactive oxygen species (ROS) and malondialdehyde (MDA) levels, and blocked the reduction of mitochondrial membrane potential (MMP) induced by H2O2. MR enhanced the activity of the antioxidant enzyme superoxide dismutase (SOD) and suppressed H2O2-induced downregulation of the antiapoptotic protein Bcl-2 and activation of the proapoptotic protein caspase-3. Finally, we found that LY294002, a specific inhibitor of the PI3K/Akt pathway, inhibited the protective effect of MR against H2O2-induced OLN-93 cell injury in the MTT and TUNEL assays. LY294002 also inhibited the expression of SOD and Bcl-2, and increased the expression of iNOS and c-caspase-3 induced by MR treatment. MR exerts protective effects against H2O2-induced OLN-93 cell injury through the PI3K/Akt signaling pathway-mediated antioxidative stress and antiapoptotic activities. MR may provide a potential strategy for SCI treatment or other related neurodegeneration.

Despite a number of studies have been carried out on TiO2 based materials as photocatalysts for water pollutant treatment, it still needs sustained effort to extend the optical range of the photocatalysts and inhibit the recombination of photo-induced carriers to improve their catalytic activities under solar light. In this work, a series of Sn-doped TiO2 with different amounts of Sn doping (1, 5, 10 and 20 mol%) were biomimetically synthesized by a facile sol–gel method using cellulosic cotton as biotemplate. The Sn-doped TiO2 materials possess a typical three-dimensional hierarchical structure of microtubes consisting of interwoven nanofibers. The photocatalytic performance was evaluated via the degradation of methylene blue (MB) (10.0 mg L−1) under Xenon lamp simulated solar irradiation. The results show that Sn(5)-TiO2 (5 mol% Sn doping) sample exhibits an outstanding photocatalytic capacity with a superior degradation rate of higher than 98% within 30 min and a good reusability without significant decrease of activity after reused for four cycles. The most significantly improved photocatalytic capacity of TiO2 is ascribed to more extra surface hydroxyl groups and accessible active sites provided by the relatively high surface area, and a higher light capturing and utilization efficiency with less recombination of the photogenerated electron-hole pairs endowed by the good synergistic effect of the special hierarchically porous microstructure and the appropriate amount of Sn doping. Whereas, the excessive Sn doping reduces the photocatalytic activity obviously, resulting from the phase transformation of TiO2 generating more rutile phase with less reactivity, the phase separation with clear grain boundary blocking the active sites, and the extra Sn4+ acting as the recombination center. This research presents a facile biomimetic synthesis strategy combined with the traditional sol–gel method to develop various ion doped metal oxides as photocatalysts with enhanced activity.

Polycyclic polyprenylated acylphloroglucinols (PPAPs) were mainly obtained from the plants of Hypericum genus of Guttiferae family, and possessed intriguing chemical structures and appealing biolog...

Nanotube can be used as a mass sensor. To design a mass sensor for evaluating a high-speed nanoparticle, in this study, we investigated the impact vibration of a cantilever nanobeam being transversally collided by a high-speed C60 at the beam's free end with an incident velocity of vIn. The capped beam contains alternately two boron nitride zones and two carbon zones on its cross section. Hence, the relaxed beam has elliptic cross section. The vibration properties were demonstrated by molecular dynamics simulation results. Beat vibration of a slim beam can be found easily. The 1st and the 2nd order natural frequencies (f1 and f2) of the beam illustrate the vibration of beam along the short and the long axes of its elliptic cross section, respectively. f2 decreases with increasing temperature. A minimal value of vIn leads to the local buckling of the beam, and a different minimal vIn leading to damage of the beam. For the same system at a specified temperature, f2 varies with vIn. When the beam bends almost uniformly, f2 decreases linearly with vIn. If vIn becomes higher, the beam has a cross section which buckles locally, and the buckling position varies during vibration. If vIn approaches the damage velocity, a fixed contraflexture point may appear on the beam due to its strong buckling. Above the damage velocity, f2 decreases sharply. These results have a potential application in design of a mass sensor.

Recently, ensemble learning paradigm has shown great potential to achieve better prediction performance in the hyperspectral image classification. However, in the traditional methods, each classifier independently searches for the optimal spectral feature subspace in series and some important features are searched repeatedly, which leads to high computing redundancy and low effective utilization of features. In this article, an evolutionary multitask ensemble learning model (EMT_EL) for hyperspectral image classification is designed. First, the model formulates the spectral feature subspaces generation into a multitask optimization problem to concurrently search for optimal feature subspaces for multiple classifiers, which would be capable to select more informative and representative feature subspaces effectively. Second, seeking the optimal feature subspace for one base classifier can assist in the optima-seeking process for some other base classifiers via sharing the useful features, which can accelerate converge toward the direction of the optimal feature subspace, avoid trapping in local optimal subspace and improve searching capability. Third, randomization-enhanced genetic operators are designed for effective and reasonable feature selection, which can facilitate the exchange of information and improve the joint searching efficiency of the feature subspace. Eventually, the quality of generated spectral feature subspaces for each base classifier is improved and the feature sharing can parse HSI data by knowing which spectral features are important. Experimental results demonstrate that the proposed method can generate the appropriate feature subspace for each base classifier, thus it has outstanding classification performance on the different hyperspectral datasets.

Image hashing embeds the image to binary codes which can boost the efficiency of approximately nearest neighbors search. F-measure is a widely-used metric for evaluating the performance of hashing methods. However, it is non-differentiable and hence it has not been used as an object function for hashing. Heuristic algorithms, e.g. evolutionary computation and particle swarm optimization (PSO), are good at optimizing non-differentiable objectives, while they are inefficient in very high-dimensional variables which are commonly used in hashing models. To address this contradict, we propose a scheme to bridge hashing methods and F-measure objective using PSO. The hashing methods are used to generate real-valued codes for images and then the parameters of quantization procedure are optimized by PSO. Our scheme can incorporate a wide range of hashing methods, heuristic optimization algorithms and non-differentiable metrics. Experimental results demonstrate that our scheme can be used to further improve the performance of existing hashing methods.

Inflammation is a major cause of neuronal injury after spinal cord injury. We hypothesized that inhibiting caspase-1 activation may reduce neuroinflammation after spinal cord injury, thus producing a protective effect in the injured spinal cord. A mouse model of T9 contusive spinal cord injury was established using an Infinite Horizon Impactor, and VX-765, a selective inhibitor of caspase-1, was administered for 7 successive days after spinal cord injury. The results showed that: (1) VX-765 inhibited spinal cord injury-induced caspase-1 activation and interleukin-1β and interleukin-18 secretion. (2) After spinal cord injury, an increase in M1 cells mainly came from local microglia rather than infiltrating macrophages. (3) Pro-inflammatory Th1Th17 cells were predominant in the Th subsets. VX-765 suppressed total macrophage infiltration, M1 macrophages/microglia, Th1 and Th1Th17 subset differentiation, and cytotoxic T cells activation; increased M2 microglia; and promoted Th2 and Treg differentiation. (4) VX-765 reduced the fibrotic area, promoted white matter myelination, alleviated motor neuron injury, and improved functional recovery. These findings suggest that VX-765 can reduce neuroinflammation and improve nerve function recovery after spinal cord injury by inhibiting caspase-1/interleukin-1β/interleukin-18. This may be a potential strategy for treating spinal cord injury. This study was approved by the Animal Care Ethics Committee of Bengbu Medical College (approval No. 2017-037) on February 23, 2017.

The appropriate combined nitrogen and phosphorus fertilization strategy is essential for obtaining higher grain yields while maintaining soil fertility. Here, a long-term split-plot design farmland experiment with five N fertilizer rates combined with four P fertilizer rates was established during 2016–2019 to determine an appropriate nitrate-N threshold in intensive managed winter wheat- summer maize cropping, and then propose the fertilization strategy based on NO3-N threshold to balances crop yield and soil nitrogen supply capacity. The results showed that N fertilizer increased accumulated NO3-N, while the combined phosphate fertilizer at each N rate reduced the accumulated NO3-N to different degrees. With the increasing of planting seasons, the residual soil NO3-N reached a steady-state balance of soil N pool when N application rate was 150–225 kg ha−1 combined 60–120 kg ha−1 P rate. The residual NO3-N threshold was determined as 100 kg ha−1 to maintain N supply capacity and prevent it leaching. Based on it, we recommend 154 kg ha−1 of N and 106 kg ha−1 of P fertilizer in the wheat season, and 162 kg ha−1 of N and 122 kg ha−1 of P fertilizer in the maize season. The optimized fertilizer strategy reduced the fertilizer by 67 kg N ha−1 per year and reduced the residual NO3-N by 34.2% in deep soil while only reducing average yield by 3.1% across crops and years. This study can serve as basis for sustainable solutions for balances grain yields and soil nitrogen supply capacity as well as preventing nitrate pollution in farmland.

Although the deep neural network (DNN) has shown a powerful ability in hyperspectral image (HSI) classification, its learning requires a large number of labeled training samples; otherwise, it is prone to over-fitting and has a poor classification performance. However, this requirement is impractical for HSIs due to the difficulty in obtaining class labels. To make DNNs suitable for HSI classification with few labeled samples, we propose a graph-based deep multitask few-shot learning (GDMFSL) framework that learns the intrinsic relationships among all samples (labeled and unlabeled) of HSIs with the assistance of graph information to alleviate the over-fitting caused by few labeled training samples. Firstly, a semi-supervised graph is constructed to generate graph information. Secondly, a deep multitask network (DMN) is designed, which contains two subnetworks (tasks): a classifier subnetwork for learning class information from labeled samples and a Siamese subnetwork for learning sample relationships from the semi-supervised graph. To effectively learn graph information, a loss function suitable for the Siamese subnetwork is designed that shortens (and expands) the distance between the target sample and its nearest (and farthest) neighbors. Finally, since the number of training samples of the two subnetworks is severely imbalanced, a multitask few-shot learning strategy is designed to make two subnetworks converge simultaneously. Experimental results on the Indian Pines, University of Pavia and Salinas datasets demonstrate that GDMFSL achieves a better classification performance relative to existing competitors in few-shot settings. In particular, when only five labels per class are involved in training, the classification accuracy of GDMFSL on the three datasets reaches 87.58%, 86.42% and 98.85%, respectively.