Search

Fast adversarial training (FAT) is an efficient method to improve robustness in white-box attack scenarios. However, the original FAT suffers from catastrophic overfitting, which dramatically and suddenly reduces robustness after a few training epochs. Although various FAT variants have been proposed to prevent overfitting, they require high training time. In this paper, we investigate the relationship between adversarial example quality and catastrophic overfitting by comparing the training processes of standard adversarial training and FAT. We find that catastrophic overfitting occurs when the attack success rate of adversarial examples becomes worse. Based on this observation, we propose a positive prior-guided adversarial initialization to prevent overfitting by improving adversarial example quality without extra training time. This initialization is generated by using high-quality adversarial perturbations from the historical training process. We provide theoretical analysis for the proposed initialization and propose a prior-guided regularization method that boosts the smoothness of the loss function. Additionally, we design a prior-guided ensemble FAT method that averages the different model weights of historical models using different decay rates. Our proposed method, called FGSM-PGK, assembles the prior-guided knowledge, i.e., the prior-guided initialization and model weights, acquired during the historical training process. The proposed method can effectively improve the model's adversarial robustness in white-box attack scenarios. Evaluations of four datasets demonstrate the superiority of the proposed method.

Digital pathology images are treated as the “gold standard” for the diagnosis of colorectal lesions, especially colon cancer. Real-time, objective and accurate inspection results will assist clinicians to choose symptomatic treatment in a timely manner, which is of great significance in clinical medicine. However, Manual methods suffers from long inspection cycle and serious reliance on subjective interpretation. It is also a challenging task for existing computer-aided diagnosis methods to obtain models that are both accurate and interpretable. Models that exhibit high accuracy are always more complex and opaque, while interpretable models may lack the necessary accuracy. Therefore, the framework of ensemble adaptive boosting prototype tree is proposed to predict the colorectal pathology images and provide interpretable inference by visualizing the decision-making process in each base learner. The results showed that the proposed method could effectively address the “accuracy-interpretability trade-off” issue by ensemble of m adaptive boosting neural prototype trees. The superior performance of the framework provides a novel paradigm for interpretable inference and high-precision prediction of pathology image patches in computational pathology.

Objective:The use of COOK®balloon in the process of induction of labor is gradually promoted. This study was conducted to investigate the safety and efficacy of COOK®double balloon in promoting the maturity of uterine neck, and whether COOK®double balloon can improve the success rate of induced labor.Methods:A total of 343 pregnant women with full-term pregnancy in Shanghai Putuo Maternity & Infant Health Hospital from January 1st to September 30th of 2016 were enrolled. Of all the pregnant women, 166 had labor induction, which included the use of a COOK®balloon, the implementation of artificial rupture of membranes (AROM) and oxytocin intravenous (IV) drip (COOK®group). The other 177 pregnant women with spontaneous rupture of membranes (SROM) and mature cervix only were administrated with oxytocin IV drip to promote regular uterus contraction (oxytocin group). Maternal adverse reactions, mode of delivery and delivery outcomes of two groups were subsequently compared between the two groups.Results:There were significant differences of maternal age (28.29±3.34versus 29.25±3.62years, P=0.02), gestational age (283.49±4.53versus 274.71±7.04years, P<0.001) and birth weight (3435.27±340.29versus 3354.63±387.96g, P=0.02) between the COOK®group and the oxytocin group. There were no significant differences in terms of gravidity, parity, mode of delivery, analgesia, 1 min and 5 min Apgar score, labor time, postpartum hemorrhage, and adverse events.Conclusion:Use of a COOK®balloon may help in full-term pregnancy for women with immature cervix and to obtain the same pregnancy results as those with mature cervix, without extra occurrence of adverse events.

Conventional defect detection and recognition algorithms for integrated circuit (IC) components, used in printed circuit boards (PCBs), are primarily based on automated optical inspection (AOI). These techniques typically extract image features using image processing models, which are heavily dependent on visual cues and can thus be inaccurate. To address these issues, we propose the use of a deep convolutional neural network, an improved LeNet-5 structure called ICChaNet with a deeper structure and more complex parameters. First, the algorithm for character extraction is implemented and used to generate character samples from IC component images. This process includes grayscaling, binarization, and contour extraction, applied to a set of IC component images to establish 53 categories of characters (letters and number with similar symbols removed). Three critical hyperparameters, including learning rate, pooling strategy, and optimization strategy, are then optimized through a comparative analysis. The effect of different network architectures on model efficiency is also investigated by varying the layer depth. A combination classifier strategy is subsequently added to the fully connected layers to strengthen the feature expression in the corresponding classes. Finally, the performance of the proposed ICChaNet model is verified through a comparison with popular CNNs. Experimental results demonstrate that ICChaNet has achieved a classification accuracy of 98.5% and a test average accuracy precision of 97.73%. These results suggest the proposed ICChaNet model is a promising approach for automated character identification.

CO2injection to enhance coalbed methane (CO2-ECBM) production is a promising technology for realizing the efficient exploitation of new energy resource, but its effect is cooperatively affected by miscellaneous influencing factors. To elucidate the mechanism and increase the efficiency of the displacement of CH4by CO2, CO2flooding experiments under the influence of the injection pressure, confining pressure, and temperature were conducted. The results show that an increase in the injection pressure enlarges the pore-fracture network of the coal matrix and promotes its gas adsorption ability and seepage capacity, while an increase in the confining pressure compresses the coal body and reduces the adsorption capacity and permeability. Increasing the temperature enhances kinetic energy and induces the thermal swelling effect, which disfavors the adsorption and seepage processes. The CO2adsorption capacity is larger than the CH4adsorption capacity due to the higher adsorption affinity of CO2, but the CO2permeability is lower than the CH4permeability due to the severe adsorption swelling effect. The CO2flooding process is characterized by three parameters (equal percentage displacement time, displacement rate, and displacement efficiency) and can be divided into the CO2driving stage, replacement and displacement stage, and equilibrium stage. The injection pressure and temperature decrease the displacement time and increase the displacement rate and displacement efficiency, while confining pressure shows the opposite trends. Multifactor interaction models of the displacement efficiency of coal seams at various buried depths were established and demonstrated to appropriately predict the displacement efficiency of CO2-ECBM in deep coal seams. The displacement efficiency is controlled by the reservoir performance, geological characteristics, and CO2injection technology. The selection of a high-permeability coal reservoir and coal seams with an appropriate buried depth and the development of a reasonable injection plan are suggested for the implementation of CO2-ECBM extraction.

Using a nondegenerate four-wave mixing (FWM) process based on a double-Λ scheme in hot cesium vapor, we demonstrate a compact diode-laser-pumped quantum light source for the generation of quantum correlated twin beams with a maximum squeezing of 6.5 dB. The squeezing is observed at a Fourier frequency in the audio band down to 0.7 kHz which, to the best of our knowledge, is the first observation of sub-kilohertz intensity-difference squeezing in an atomic system so far. A phase-matching condition is also investigated in our system, which confirms the spatial-multi-mode characteristics of the FWM process. Our compact low-frequency squeezed light source may find applications in quantum imaging, quantum metrology, and the transfer of optical squeezing onto a matter wave.

[Display omitted]

Locking plate (LP) re-fixation is mainly used to treat postoperative implant periprosthetic refractures; however, the extensive trauma and the fixation form of LP make the operation difficult. The bridge combined fixation system (BCFS) is a new clip-rod internal fixation system, and its clinical application is in its infancy. To compare the clinical effect of BCFS and LP in the treatment of geriatric postoperative implant periprosthetic refracture following proximal femoral fracture surgery. Thirty-two patients (14 with BCFS and 18 with LP) with postoperative implant periprosthetic refracture following proximal femoral fracture surgery, who underwent surgery in our hospital, were analyzed retrospectively. The incision length, operation time, intraoperative bleeding volume, postoperative drainage volume, postoperative hospital stay, fracture healing time and complications of each patient were recorded. Regular radiographs were taken after the operation to evaluate the fracture reduction and fixation. All the patients were followed for 12 months to evaluate their limb function by Johner‐Wruhs scoring criteria. The patients were followed for an average of 24.1 months, and all achieved bony union, with no complications such as infection, nonunion, and internal fixation instrument falling off and loosening after the operation. Delayed healing occurred in two cases in the LP group. The average value of surgical incision length, operation time, postoperative hospitalization time and fracture healing time in the BCFS group were significantly smaller than those in the LP group, accompanied by a decrease in intraoperative bleeding and postoperative drainage volumes (P < .05). The rate of limb function in the BCFS group (85.7%) was higher than that in the LP group (83.3%), with no significance (P > .05). The BCFS in the refracture around the implant of the proximal femoral fracture exhibited many advantages such as simple operation, strong plasticity, effective reduction of surgical trauma, promotion of fracture healing and early functional rehabilitation, etc, making it an advantageous clinical application.

A polarization squeezed optical beam can directly interact with the atomic medium, and its measurement is local-oscillator free. Here we have deterministically generated polarization squeezed light at 795 nm. First, a laser at 398 nm is produced from second-harmonic generation with an enhancement cavity. Then a pair of quadrature amplitude squeezed optical fields is prepared with two degenerate optical parameter amplifiers that are pumped by the resulting laser at 398 nm. Finally, the polarization squeezed state of light is generated by combining the above two quadrature amplitude squeezed optical beams on a polarizing beam splitter, and the quantum fluctuations of three Stokes operators simultaneously are reduced 4.0 dB below the quantum noise limit. The polarization squeezing has potential applications in future quantum information networks.

The direct generation of discrete-variable three-photon polarization entanglement has been recently demonstrated [Nat. Photonics8, 801 (2014)10.1038/nphoton.2014.218NPAHBY1749-4885]. Here, we propose a feasible scheme to generate continuous-variable (CV) three-color polarization entangled states of light in a deterministic way. The method initially prepares a tripartite quadrature entanglement based on two cascaded nondegenerate optical parametric oscillators. The entangled modes are then coupled, respectively, with three strong coherent optical beams on polarized beam splitters, leading to the generation of the three-color CV polarization entanglement. The polarization entanglement is then verified by inseparability criterion and positive partial transposition criterion. Additionally, optimal parameters in practical conditions are obtained through numerical simulations, which could provide useful guidance for experimental implementations. The demonstrated CV three-color polarization entanglement is suitable in quantum information processing tasks relying on the direct interaction between spin of atoms and polarization of light.

The preparation of multipartite entangled states is the prerequisite for exploring quantum information networks and quantum computation. In this paper, we review the experimental progress in the preparation of cluster states and multi-color entangled states with continuous variables. The preparation of lager scale multipartite entangled state provide valuable quantum resources to implement more complex quantum informational tasks.

Aerobic composting of cellulosic biomass is difficult to start at low temperatures. Especially, the effect of microbial inoculants on the initiation of aerobic composting of cellulosic material at low temperature and the succession of microbial community is still poorly defined. In this study, the aerobic composting of mushroom residue and wood chips successfully started at 3–11 °C by inoculating the microbial agents (including bacteria and fungi) for cellulose degradation and low temperature adaptation. It was found that the inoculum was beneficial for the heating of the compost pile, starting of the aerobic composting process and the degradation of cellulose. The thermophilic stage of inoculation group lasted for 13 days and 8 days longer than that of control group. The degradation efficiencies of hemicellulose and cellulose in inoculation group and control group were 37.50% and 27.30%, while the degradation efficiencies of cellulose were 21.29% and 16.18%, respectively. Furthermore, the results of high-throughput sequencing revealed that inoculation of microbial agents was beneficial to the optimization of community structure and the conversion of cellulose substrate. In addition, the inoculated strains of Aspergillus, Penicillium, Bacillusand Streptomycesmight play active roles in the composting process, especially in the set-up and thermophilic stage of the composting.

Pattern encoding and decoding are two challenging problems in a three-dimensional (3D) reconstruction system using coded structured light (CSL). In this paper, a one-shot pattern is designed as an M-array with eight embedded geometric shapes, in which each 2 × 2 subwindow appears only once. A robust pattern decoding method for reconstructing objects from a one-shot pattern is then proposed. The decoding approach relies on the robust pattern element tracking algorithm (PETA) and generic features of pattern elements to segment and cluster the projected structured light pattern from a single captured image. A deep convolution neural network (DCNN) and chain sequence features are used to accurately classify pattern elements and key points (KPs), respectively. Meanwhile, a training dataset is established, which contains many pattern elements with various blur levels and distortions. Experimental results show that the proposed approach can be used to reconstruct 3D objects.

The preparation of multipartite entangled states is the prerequisite for exploring quantum information networks and quantum computation. In this Letter, we present what we believe is the first experimental demonstration of an eight-partite spatially separated continuous variable (CV) cluster state of optical modes. Via the linearly optical transformation of eight squeezed states of light, the eight-partite cluster entangled state with amplitude and phase quadrature correlations are prepared. The generated eight entangled photonic qumodes are spatially separated, which provides valuable quantum resources for implementing CV quantum information protocols.

We experimentally demonstrate that the quantum entanglement between amplitude and phase quadratures of optical modes produced from a nondegenerate optical parametric amplifier (NOPA) can be enhanced and manipulated phase sensitively by means of another NOPA. When both NOPAs operate at deamplification, the entanglement degree is increased at the cavity resonance of the second NOPA. When the first NOPA operates at deamplification and the second one at amplification, the spectral features of the correlation variances are significantly changed. The experimental results are in good agreement with the theoretical expectation.

Quantum network enables quantum communication among quantum nodes and provides advantages that are unavailable in any classical network. Based on rapidly developing science and technology in quantum communication, the studies on quantum network have also made important progresses recent years. In this study, we briefly review the experimental progresses in building quantum network based on optical field and discuss the challenges toward a quantum Internet.

The quantum entanglement of amplitude and phase quadratures between two intense optical beams with a total intensity of 22mW and a frequency difference of 1nm, which are produced from an optical parametric oscillator operating above threshold, is experimentally demonstrated with two sets of unbalanced Mach-Zehnder interferometers. The measured quantum correlations of intensity and phase are in reasonable agreement with the results calculated based on a semiclassical analysis of the noise characteristics given by Fabre [J. Phys. (France) 50, 1209 (1989)].

We investigate the different characteristics of the intensity noise of a laser-diode-pumped single-frequency ring Nd:YVO _4 laser and a Nd:YVO _4–KTP green laser. By use of an optoelectronic feedback circuit connected directly to the pump current of the laser diode, the low-frequency intensity noise of the intracavity frequency doubler was suppressed to some extent.

We investigate the different characteristics of the intensity noise of a laser-diode-pumped single-frequency ring Nd:YVO_4 laser and a Nd:YVO_4–KTP green laser. By use of an optoelectronic feedback circuit connected directly to the pump current of the laser diode, the low-frequency intensity noise of the intracavity frequency doubler was suppressed to some extent.

We experimentally present a high-efficiency generation of low-noise 447 nm laser by frequency doubling of a Ti:sapphire laser in an external ring cavity with a PPKTP crystal. The maximum blue laser power of 308 mW has been obtained and the corresponding conversion efficiency can reach 70%, when 440 mW infrared laser is injected. Moreover, the noise of the resulting blue laser can reach the quantum noise limit when the analyzing frequency is above 1.5 MHz. Our system provides an ideal pump field for an optical parametric amplifier for the generation of non-classical states of light matching matter absorption line.

The intensity difference squeezed state, which means that the fluctuation of the intensity difference between signal and idler beams is less than that of the corresponding shot noise level (SNL), plays an important role in high sensitivity measurement, quantum imaging, and quantum random numbers generation. When an optical parametric oscillator consisting of a type-II phase-matching periodically poled KTiOPO_4 crystal operates above the threshold, an intensity difference squeezed state at a telecommunication wavelength can be obtained. The squeezing of 7.7±0.5??dB below the SNL is achieved in an analysis frequency region of 2.4–5.0??MHz.

Frequency tunable continuous variable (CV) entangled optical beams are experimentally demonstrated from a non-degenerate optical parametric oscillator working above the threshold. The measured correlation variances of amplitude and phase quadratures are 3.2 and 1.5 dB, respectively, below the corresponding shot noise level (SNL) in the tuning range of 580 GHz (2.25 nm). The frequency tuning is realized by simply controlling the temperature of the nonlinear crystal.