PSE-Net: Channel pruning for Convolutional Neural Networks with parallel-subnets estimator.

Channel Pruning is one of the most widespread techniques used to compress deep neural networks while maintaining their performances. Currently, a typical pruning algorithm leverages neural architecture search to directly find networks with a configurable width, the key step of which is to identify representative subnet for various pruning ratios by training a supernet. However, current methods mainly follow a serial training strategy to optimize supernet, which is very time-consuming. In this work, we introduce PSE-Net, a novel parallel-subnets estimator for efficient channel pruning. Specifically, we propose a parallel-subnets training algorithm that simulate the forward-backward pass of multiple subnets by droping extraneous features on batch dimension, thus various subnets could be trained in one round. Our proposed algorithm facilitates the efficiency of supernet training and equips the network with the ability to interpolate the accuracy of unsampled subnets, enabling PSE-Net to effectively evaluate and rank the subnets. Over the trained supernet, we develop a prior-distributed-based sampling algorithm to boost the performance of classical evolutionary search. Such algorithm utilizes the prior information of supernet training phase to assist in the search of optimal subnets while tackling the challenge of discovering samples that satisfy resource constraints due to the long-tail distribution of network configuration. Extensive experiments demonstrate PSE-Net outperforms previous state-of-the-art channel pruning methods on the ImageNet dataset while retaining superior supernet training efficiency. For example, under 300M FLOPs constraint, our pruned MobileNetV2 achieves 75.2% Top-1 accuracy on ImageNet dataset, exceeding the original MobileNetV2 by 2.6 units while only cost 30%/16% times than BCNet/AutoAlim.

Hypopituitary syndrome with pituitary crisis in a patient with traumatic shock: A case report.

BACKGROUND: Shock is among the most common conditions that clinicians face in intensive care unit (ICU), of which hypovolemic shock is encountered most frequently; some patients instead suffer from neurogenic, cardiogenic, or infectious forms of shock. However, there are additional types of shock from unusual causes that are often undiagnosed. Here, we report the case of a patient who was initially misdiagnosed with hypovolemic shock, but exhibited persistent hypotension because of continuous fluid replacement and vasoactive drug administration, and was eventually diagnosed with hypopituitarism with crisis. CASE SUMMARY: A 73-year-old Chinese man was admitted to the neurosurgery department following injury caused by a heavy object with symptoms of anemia and high fever. He was transferred to the ICU on the fourth day after hospitalization because of hypotension and unconsciousness. Blood analysis indicated that the patient was suffering from anemia and thrombocytopenia. Ultrasonography showed that there was no apparent abnormality in the cardiac structure but there was mild tricuspid regurgitation. Computed tomography revealed that there were signs of hemorrhage at the right basal ganglia; accordingly, hypovolemic shock, possibly septic shock, was initially considered. Even after routine treatment for shock, the hypotension remained severe. The patient was again thoroughly examined to investigate the underlying cause. The antishock therapy was supplemented with corticosteroids to counter potential hypopituitarism. The patient made a full recovery, and the blood pressure returned to normal. CONCLUSION: A case of pituitary adenoma with multiple injuries was identified. Because of hypopituitarism, functionality of the corresponding endocrine system was restricted, with the most pronounced manifestation being unstable blood circulation requiring hormone replacement therapy. Such cases are relatively rare but may occur if multiple injuries are sustained. The present case represents a reference for the clinical treatment of patients with multiple injuries.

Hybrid Deep Recurrent Neural Networks for Noise Reduction of MEMS-IMU with Static and Dynamic Conditions.

Micro-electro-mechanical system inertial measurement unit (MEMS-IMU), a core component in many navigation systems, directly determines the accuracy of inertial navigation system; however, MEMS-IMU system is often affected by various factors such as environmental noise, electronic noise, mechanical noise and manufacturing error. These can seriously affect the application of MEMS-IMU used in different fields. Focus has been on MEMS gyro since it is an essential and, yet, complex sensor in MEMS-IMU which is very sensitive to noises and errors from the random sources. In this study, recurrent neural networks are hybridized in four different ways for noise reduction and accuracy improvement in MEMS gyro. These are two-layer homogenous recurrent networks built on long short term memory (LSTM-LSTM) and gated recurrent unit (GRU-GRU), respectively; and another two-layer but heterogeneous deep networks built on long short term memory-gated recurrent unit (LSTM-GRU) and a gated recurrent unit-long short term memory (GRU-LSTM). Practical implementation with static and dynamic experiments was carried out for a custom MEMS-IMU to validate the proposed networks, and the results show that GRU-LSTM seems to be overfitting large amount data testing for three-dimensional axis gyro in the static test. However, for X-axis and Y-axis gyro, LSTM-GRU had the best noise reduction effect with over 90% improvement in the three axes. For Z-axis gyroscope, LSTM-GRU performed better than LSTM-LSTM and GRU-GRU in quantization noise and angular random walk, while LSTM-LSTM shows better improvement than both GRU-GRU and LSTM-GRU networks in terms of zero bias stability. In the dynamic experiments, the Hilbert spectrum carried out revealed that time-frequency energy of the LSTM-LSTM, GRU-GRU, and GRU-LSTM denoising are higher compared to LSTM-GRU in terms of the whole frequency domain. Similarly, Allan variance analysis also shows that LSTM-GRU has a better denoising effect than the other networks in the dynamic experiments. Overall, the experimental results demonstrate the effectiveness of deep learning algorithms in MEMS gyro noise reduction, among which LSTM-GRU network shows the best noise reduction effect and great potential for application in the MEMS gyroscope area.

A neonatal murine model of coxsackievirus A4 infection for evaluation of vaccines and antiviral drugs.

Coxsackievirus A4 (CVA4) infection can cause hand, foot and mouth disease (HFMD), an epidemic illness affecting neonatal and paediatric cohorts, which can develop to severe neurological disease with high mortality. In this study, we established the first ICR mouse model of CVA4 infection for the evaluation of inactivated vaccines and antiviral drug screening. The CVA4 YT226R strain was selected to infect the neonatal mice and three infectious factors were optimized to establish the infection model. The 3-day-old neonatal mice exhibited clinical symptoms such as hind limb paralysis and death. The severe inflammatory reactions were closely related to the abnormal expression of the acute phase response proinflammatory cytokine IL-6 and an imbalance in the IFN-γ/IL-4 ratio. Importantly, the inactivated CVA4 whole-virus vaccine induced humoral immune responses in adult females and the maternal antibodies afforded mice complete protection against lethal dose challenges of homologous or heterologous CVA4 strains. Both IFN-α2a and antiserum inhibited the replication of CVA4 and increased the survival rates of neonatal mice during the early stages of infection. This neonatal murine model of CVA4 infection will be useful for the development of prophylactic and therapeutic vaccines and for screening of antiviral drugs targeting CVA4 to decrease morbidity and mortality.

Characterization and evolution of gene clusters for terpenoid phytoalexin biosynthesis in tobacco.

MAIN CONCLUSION: The study performed genome-wide identification, characterization and evolution analysis of gene clusters for phytoalexin terpenoid biosynthesis in tobacco, and specifically illustrated ones for capsidiol, an efficient defensive specialized metabolite. Terpenoid phytoalexins play an important role in plant self-defense against pest and pathogen attack. Terpenoid biosynthesis involves terpene synthase and cytochrome P450, which always locate and function as cluster(s). In this study, we performed genome-wide investigation of metabolic gene clusters involved in terpenoid production in tobacco (Nicotiana tabacum). Due to the complexity of the tobacco genome, we modified a published prediction pipeline to reduce the influence of the large number of repeats and to improve the annotation of tobacco genes with respect to their metabolic functions. We identified 1181 metabolic gene clusters with 34 of them potentially being involved in terpenoid biosynthesis. Through integration with transcriptome and metabolic pathway annotation analyses, 3 of the 34 terpenoid biosynthesis-related gene clusters were determined to be high-confidence ones, with 2 involved in biosynthesis of capsidiol, a terpenoid recognized as 1 of the effective resistance compounds in the Nicotiana species. The capsidiol-related gene cluster was conserved in N. sylvestris, N. tomentosiformis and N. attenuate. Our findings demonstrate that phytoalexins in tobacco can arise from operon-like gene clusters, a genomic pattern characterized as being beneficial for rapid stress response, gene co-regulation, co-function and co-heredity.

Effects of Acetylshikonin on the Infection and Replication of Coxsackievirus A16 in Vitro and in Vivo.

Coxsackievirus A16 (CVA16) is one of the most prevalent enteroviral pathogens associated with hand, foot, and mouth disease. In the present study, we have investigated (1) whether the bioactive compound acetylshikonin (AS) inhibits CVA16 infection in vitro and in vivo and (2) the potential antiviral mechanism(s). The results suggest that AS is nontoxic at concentrations of up to 5 µmol/L and could directly inactivate virus particles at relatively low concentrations (0.08 µmol/L), thereby rendering CVA16 incapable of cellular entry. Correspondingly, the expression of viral RNA in vitro was also reduced 100-fold ( P < 0.05) when compared to infected, untreated controls. Results from a CVA16-infected neonatal mouse model indicate that, in comparison to the virus-infected, untreated group, body weights of the mice in the virus-infected, compound-treated group increased more steadily with less severe clinical symptoms. In addition, viral loads in internal organs significantly decreased in treated animals, concomitantly with both reduced pathology and diminished expression of the proinflammatory cytokines IFN-γ and IL-6. In conclusion, AS exerted an inhibitory effect on CVA16 infection in vitro and in vivo. Our study provides a basis for further investigations of AS-type compounds to develop therapeutics to mitigate CVA-associated disease in children.

Genome Analysis of Coxsackievirus A4 Isolates From Hand, Foot, and Mouth Disease Cases in Shandong, China.

Coxsackievirus A4 (CVA4) is one of the most prevalent pathogens associated with hand, foot and mouth disease (HFMD), an acute febrile illness in children, and is also associated with acute localized exanthema, myocarditis, hepatitis and pancreatitis. Despite this, limited CVA4 genome sequences are currently available. Herein, complete genome sequences from CVA4 strains (n = 21), isolated from patients with HFMD in Shandong province, China between 2014 and 2016, were determined and phylogenetically characterized. Phylogenetic analysis of the VP1 gene from a larger CVA4 collection (n = 175) showed that CVA4 has evolved into four separable genotypes: A, B, C, and D; and genotype D could be further classified in to two sub-genotypes: D1 and D2. Each of the 21 newly described genomes derived from isolates that segregated with sub-genotype D2. The CVA4 genomes displayed significant intra-genotypic genetic diversity with frequent synonymous substitutions occurring at the third codon positions, particularly within the P2 region. However, VP1 was relatively stable and therefore represents a potential target for molecular diagnostics assays and also for the rational design of vaccine epitopes. The substitution rate of VP1 was estimated to be 5.12 × 10-3 substitutions/site/year, indicative of ongoing CVA4 evolution. Mutations at amino acid residue 169 in VP1 gene may be responsible for differing virulence of CVA4 strains. Bayesian skyline plot analysis showed that the population size of CVA4 has experienced several dynamic fluctuations since 1948. In summary, we describe the phylogenetic and molecular characterization of 21 complete genomes from CVA4 isolates which greatly enriches the known genomic diversity of CVA4 and underscores the need for further surveillance of CVA4 in China.

2-Fluoro-1,3-diamino-4,6-dinitrobenzene (ZXC-7) and 2-Fluoro-1,3,5-triamino-4,6-dinitrobenzene (ZXC-8): Thermally Stable Explosives with Outstanding Properties.

Two novel, thermally stable explosives, 2-fluoro-1,3-diamino-4,6-dinitrobenzene (ZXC-7) and 2-fluoro-1,3,5-triamino-4,6-dinitrobenzene (ZXC-8), are reported. These two compounds can be prepared by means of simple synthetic methods and they show outstanding properties (detonation velocity, detonation pressure, sensitivity toward mechanical stimuli, and temperature of decomposition). Their structures are very similar to that of 1,3,5-triamino-2,4,6-trinitrobenzene (TATB). They were isolated and characterized by means of mass spectrometry and multinuclear (1 H, 13 C) NMR spectroscopy. The structure of ZXC-7 in the crystalline state was determined by low-temperature single-crystal X-ray diffraction. From the calculated standard molar enthalpy of formation (CBS-4 M) and the densities (which were determined by a gas pycnometer), the Chapman-Jouguet detonation properties were predicted by using the EXPLO5 V6.01 thermochemical computer code. The sensitivities of ZXC-7 and ZXC-8 towards impact and friction were determined.

Synthesis and Detonation Properties of 5-Amino-2,4,6-trinitro-1,3-dihydroxy-benzene.

5-Amino-4,6-dinitro-1,3-dihydroxy-benzene (6) was synthesized through the ring-opening reaction of macrocyclic compound 4 with the aid of VNS (vicarious nucleophilic substitution of hydrogen) reaction conditions. The mechanism of ring opening of macrocyclic compound 4 was studied. 5-Amino-2,4,6-trinitro-1,3-dihydroxy-benzene (8) was obtained after the nitration of 6 in KNO3 and concentrated sulfuric acid. The thermal stability, sensitivity, and other detonation performances of 6 or 8 were compared to commercially used 1,3,5-triamino-2,4,6-trinitrobenzene (TATB) or 1,3,5-trinitrotriazacyclohexane (RDX), respectively. All target compounds were characterized by using single-crystal X-ray diffraction, NMR spectroscopy, elemental analysis, and differential scanning calorimetry. The sensitivities were determined by using BAM methods (drop-hammer and friction tests). Performance parameters, including heats of formation and detonation properties, were calculated by using Gaussian 03 and EXPLO5 v6.01 programs, respectively. It is worth pointing out that compound 8 has a remarkable measured density of 2.078 g cm-3 at 298 K. In addition, compound 8 is more insensitive than RDX (compound 8: IS=11 J; RDX: IS=7 J; IS is the impact sensitivity).