Search

Click-through rate (CTR) prediction plays an important role in many industrial applications, feature engineering directly influences CTR prediction performance because features are normally the multi-field type. However, the existing CTR prediction techniques either neglect the importance of each feature or regard the feature interactions equally for feature learning. In addition, using an inner product or a Hadamard product is too simple to effectively model the feature interactions. These limitations lead to suboptimal performances of existing models. In this paper, we propose a framework called Hierarchical Attention and Feature Projection neural network (HAFP) for CTR prediction, which enables the automatically learning of more representative and efficient feature representation in an end-to-end manner. Towards this end, we employ a feature learning layer with a hierarchical attention mechanism to jointly extract more generalized and dominant features and feature interactions. In addition, a projective bilinear function is designed in meaningful second-order interaction encoder to effectively learn more fine-grained and comprehensive second-order feature interactions. Taking advantages of the hierarchical attention mechanism and the projective bilinear function, our proposed model can not only model feature learning in a flexible fashion, but also provide an interpretable capability of the prediction results. Experimental results on two real-world datasets demonstrate that HAFP outperforms the state-of-the-art in terms of Logloss and AUC for CTR prediction baselines. Further analysis verifies the importance of the proposed hierarchical attention mechanism and the projective bilinear function for modelling the feature representation, showing the rationality and effectiveness of HAFP.

The task of session-based recommendation is predicting the next recommendation item when available information only includes the anonymous behavior sequence. Previous methods of session-based recommendation usually integrate the general interest, dynamic interest, and current interest to promote recommendation performance. However, most existing methods ignore the non-monotone feature interactions when building user’s dynamic interest and model item-item transitions through a linear way when building user’s current interest, which reduces the performance of model. In this paper, we design a novel method for session-based recommendation with recurrent and convolutional neural network. Specifically, The Gated Recurrent Unit with item-level attention mechanism learns the user’s general interest, while the convolutional operation with horizontal filter and vertical filter search for user’s current interest and dynamic interest. Moreover, the outputs of recurrent operation and convolutional operation are concatenated to generate the recommendation. Furthermore, we evaluate the proposed model on three real-world datasets which come from e-commerce and music API, respectively. The experimental results show that our model outperforms the state-of-the-art methods on session-based recommendation.

Land cover segmentation is an important and challenging task in the field of remote sensing. Even though convolutional neural networks (CNNs) provide great support for semantic segmentation, standard models are still difficult to capture global information and long-range dependencies in remote sensing images. To overcome these limitations, we proposed an attention guided encoder-decoder network with multi-scale context aggregation to achieve more accurate segmentation of land cover. Based on the structure of the encoder-decoder network, we introduce a multi-scale feature fusion module with two attention modules to the top of the encoder. The multi-scale feature fusion module is employed to aggregate multi-scale features and capture global correlations. The attention modules are used to exploit the long-range dependencies and the interdependence between channels from the perspective of space and channel respectively. The experimental results on the GF-2 images show that our proposed method achieves state-of-the-art performance, with an OA of 84.1% and the mIoU of 62.3%. Compared with the baseline network, our method improves the OA by 3.3% and the mIoU by 4.4%. The comparative experiments also demonstrate that the proposed approach can significantly improve the accuracy of land cover segmentation than other compared methods.

We report on the generation of 7.6 ns, 162 μJ and 21.3 kW (peak power) single-frequency optical pulses at 780 nm with near-diffraction limited beam quality (M 2

In scene-level classification of remote sensing, fusion of multi-feature can significantly boost the performance. However, most methods directly fuse the features of different modalities wi...

When convolutional neural networks (CNNs) are applied to the synthetic aperture radar (SAR) image classification, they are prone to overfitting due to scarce SAR image data, and CNNs require a large amount of storage and long computing time, so it is difficult to deploy them on resource constrained devices. This letter proposes a simple and feasible approach that can effectively solve these problems. First, the convolutional layers of the pretrained model on the ImageNet data set are transferred, and a new convolutional layer and global pooling layer are added afterward. Then, fine-tuning is performed on the new network from the SAR image data set. Finally, a filter-based pruning method is used on the convolutional layers to obtain a compact network. Compared with the all-convolutional network (A-ConvNets) which is the state-of-the-art method on the moving and stationary target acquisition and recognition data set, our method achieves about $3.6\times $ speedup during forward propagation and $3.7\times $ compression of the parameters, with only a 1.42% decrease in the accuracy.

Unsupervised domain adaptation (UDA) based on adversarial learning for remote-sensing scene classification has become a research hotspot because of the need to alleviating the lack of annotated training data. Existing methods train classifiers according to their ability to distinguish features from source or target domains. However, they suffer from the following two limitations: (1) the classifier is trained on source samples and forms a source-domain-specific boundary, which ignores features from the target domain and (2) semantically meaningful features are merely built from the adversary of a generator and a discriminator, which ignore selecting the domain invariant features. These issues limit the distribution matching performance of source and target domains, since each domain has its distinctive characteristic. To resolve these issues, we propose a framework with error-correcting boundaries and feature adaptation metric. Specifically, we design an error-correcting boundaries mechanism to build target-domain-specific classifier boundaries via multi-classifiers and error-correcting discrepancy loss, which significantly distinguish target samples and reduce their distinguished uncertainty. Then, we employ a feature adaptation metric structure to enhance the adaptation of ambiguous features via shallow layers of the backbone convolutional neural network and alignment loss, which automatically learns domain invariant features. The experimental results on four public datasets outperform other UDA methods of remote-sensing scene classification.

Recent efforts have focused on customizing orthobiologics, such as platelet-rich plasma (PRP) and bone marrow concentrate (BMC), to improve tissue repair. We hypothesized that oral losartan (a TGF-&beta, 1 blocker with anti-fibrotic properties) could decrease TGF-&beta, 1 levels in leukocyte-poor PRP (LP-PRP) and fibrocytes in BMC. Ten rabbits were randomized into two groups (N = 5/group): osteochondral defect + microfracture (control, group 1) and osteochondral defect + microfracture + losartan (losartan, group 2). For group 2, a dose of 10mg/kg/day of losartan was administrated orally for 12 weeks post-operatively. After 12 weeks, whole blood (WB) and bone marrow aspirate (BMA) samples were collected to process LP-PRP and BMC. TGF-&beta, 1 concentrations were measured in WB and LP-PRP with multiplex immunoassay. BMC cell populations were analyzed by flow cytometry with CD31, CD44, CD45, CD34, CD146 and CD90 antibodies. There was no significant difference in TGF-&beta, 1 levels between the losartan and control group in WB or LP-PRP. In BMC, the percentage of CD31+ cells (endothelial cells) in the losartan group was significantly higher than the control group (p = 0.008), while the percentage of CD45+ cells (hematopoietic cells-fibrocytes) in the losartan group was significantly lower than the control group (p = 0.03).

Owing to the complexity of spatial and structural patterns of remote sensing images, scene classification of them is still an open problem and remains active in the community. The inadequacy of labeled samples leads to low accuracy of remote sensing image scene classification. To solve this problem, a classification method DCNN_MSFF is proposed based on deep convolutional neural networks (DCNN) and multi-scale features fusion (MSFF). Firstly, the remote sensing images are transformed to obtain a number of different scale ones for augmentation. Then, they are input into the DCNN for features extraction, and the different scale features of the convolutional and the fully-connected layers are encoded or pooled averagely. Finally, the processed features are fused, and the multi-kernel support vector machine (MKSVM) is used to classify the scenes. The test results in the commonly used remote sensing datasets show that, this proposed method outperforms the state-of-the-art ones in the scene classification of remote sensing images. In this paper, the multi-scale images and features of the convolutional and the fully-connected layers in the deep learning process are utilized to enhance the representation abilities of the classification features. At the same time, the MKSVM is used to improve the generalization ability of the fusion features, so the classification result is better.

The selection of scene matching areas is a difficult problem in the field of matching guidance. Compared with the traditional methods of matching feature extraction and pattern classification, this letter applies convolutional neural networks (CNN) to the extraction of synthetic aperture radar (SAR) scene matching regions for the first time. First of all, we match the SAR images of the same land taken by satellites from different angles and in different phases, and then automatically label the matching suitability of the images as the output of the network according to the matching results. Next, the digital elevation model data reflecting the elevation information and the SAR image grayscale information are fused as the input to the network. Finally, CNN is used to automatically extract the matching features and classify the suitability of the SAR images. The proposed method avoids the steps of extracting features manually and improves the classification performance of SAR scene matching area. Compared with the support vector machine method, the classification accuracy increases from 86.1% to 93.3%.

Objective To determine the clinical features of confirmed cases of severe fever with thrombocytopenia syndrome (SFTS) and to explore the early warning indicators of death from SFTS. Methods A retrospective case–control study was performed at a single medical institution in Yantai. A total of 20 SFTS patients who died (death group) during January 2014 to December 2015 and another 40 age- and sex-matched SFTS patients who survived (survivor group) were identified from the case records. The differences in demographic characteristics, clinical signs and symptoms, and laboratory parameters in the early stage of disease were compared between the two groups. Conditional logistic regression was used to identify the independent risk factors for mortality in SFTS patients. Results Univariate logistic regression analysis showed that a disturbance of consciousness, pulse–temperature deficit, neurological signs, hemorrhagic manifestations, pulmonary infection, decreased lymphocyte percentage, high lactate dehydrogenase and creatine kinase levels, increased serum creatinine, blood urea nitrogen, and C-reactive protein (CRP), hyponatremia, and prolonged activated partial thromboplastin time (APPT) and prothrombin time were associated with mortality. On multivariate logistic regression analysis, the independent predictors of death were neurological signs (odds ratio (OR) 31.247, 95% confidence interval (CI) 4.813–202.853), hemorrhagic manifestations (OR 20.251, 95% CI 2.056–199.443), disturbance of consciousness (OR 15.359, 95% CI 2.139–110.268), hyponatremia (OR 5.280, 95% CI 1.235–22.575), increased CRP (OR 2.641, 95% CI 1.090–6.396), increased serum creatinine (OR 6.776, 95% CI 1.047–43.840), and prolonged APTT (OR 6.018, 95% CI 1.450–24.975). Conclusions Neurological signs, hemorrhagic manifestations, disturbance of consciousness, hyponatremia, prolonged APTT, and increased CRP and serum creatinine are risk factors for death in SFTS.

Scene classification of high-resolution remote sensing (HRRS) image is an important research topic and has been applied broadly in many fields. Deep learning method has shown its high potential to in this domain, owing to its powerful learning ability of characterizing complex patterns. However the deep learning methods omit some global and local information of the HRRS image. To this end, in this article we show efforts to adopt explicit global and local information to provide complementary information to deep models. Specifically, we use a patch based MS-CLBP method to acquire global and local representations, and then we consider a pretrained CNN model as a feature extractor and extract deep hierarchical features from full-connection layers. After fisher vector (FV) encoding, we obtain the holistic visual representation of the scene image. We view the scene classification as a reconstruction procedure and train several class-specific stack denoising autoencoders (SDAEs) of corresponding class, i.e., one SDAE per class, and classify the test image according to the reconstruction error. Experimental results show that our combination method outperforms the state-of-the-art deep learning classification methods without employing fine-tuning.

Sequential recommendation (SR) aims to recommend items based on user information and behavior sequences. Almost all the existing works for SR construct short-term preference and long-term preference only based on the user–item interactions or the reviews rather than considering the two types of information simultaneously. In fact, interaction items and reviews commonly reflect the user’s semantic information, and play significant roles in modeling the user preference. In this paper, we propose a novel model named Parallel Item sequential pattern and Review Sequential Pattern (PIRSP) for the sequential recommendation. Specifically, first, PIRSP learns two levels of sequential patterns from item and review information, respectively: (1) item sequential pattern, which uses a gated recurrent unit with an item-attention mechanism to model history behavior sequences; (2) review sequential pattern, which takes a convolution neural network with a target-attention mechanism for modeling associated reviews of interaction items. Then, we introduce a fusion gating mechanism for selectively combining the two sequential patterns to learn the short-term preference. Second, we employ a convolution neural network with aspect information to learn the long-term preference. Finally, we utilize a linear fusion on the long-term preference and short-term preference for modeling user preference and making final recommendation. The experimental results indicate that our model outperforms other state-of-the-art methods on the Amazon dataset. Our analysis of PIRSP’s recommendation process shows the positive effect of the two types of information and fusion gating mechanism on the performance of sequential recommendation.

It is time consuming to annotate unlabeled remote sensing images. One strategy is taking the labeled remote sensing images from another domain as training samples, and the target remote sensing labels are predicted by supervised classification. However, this may lead to negative transfer due to the distribution difference between the two domains. To address this issue, we propose a novel domain adaptation method through transferring the parameters of extreme learning machine (ELM). The core of this method is learning a transformation to map the target ELM parameters to the source, making the classifier parameters of the target domain maximally aligned with the source. Our method has several advantages which was previously unavailable within a single method: multiclass adaptation through parameter transferring, learning the final classifier and transformation simultaneously, and avoiding negative transfer. We perform experiments on three data sets that indicate improved accuracy and computational advantages compared to baseline approaches.

Bladder cancer (BLCA) is one of the most malignant cancers worldwide, and its prognosis varies. 1214 BLCA samples in five different datasets and 2 platforms were enrolled in this study. By utilizing the gene expression in The Cancer Genome Atlas (TCGA) dataset, and another two datasets, in GSE13507 and GSE31684, we constructed a risk score staging system with Cox multivariate regression to evaluate predict the outcome of BLCA patients. Three genes consist of RCOR1, ST3GAL5, and COL10A1 were used to predict the survival of BLCA patients. The patients with low risk score have a better survival rate than those with high risk score, significantly. The survival profiles of another two datasets (GSE13507 and GSE31684), which were used for candidate gene selection, were similar as the training dataset (TCGA). Furthermore, survival prediction effect of risk score staging system in another 2 independent datasets, GSE40875 and E-TABM-4321, were also validated. Compared with other clinical observations, and the risk score performs better in evaluating the survival of BLCA patients. Moreover, the correlation between radiation were also evaluated, and we found that patients have a poor survival in high risk group, regardless of radiation. Gene Set Enrichment Analysis was also implemented to find the difference between high-risk and low-risk groups on biological pathways, and focal adhesion and JAK signaling pathway were significantly enriched. In summary, we developed a risk staging model for BLCA patients with three gene expression. The model is independent from and performs better than other clinical information.

In view of the problems of exposure bias and insufficient annotation data in the existing Chinese short text summary algorithms, a new pre-training short text summary algorithm named ERNIE-GEN-CTS based on ERNIE-GEN is proposed. This paper uses pre-training language model to realize the double attention of word level and word level for short text, and fills the sub block with semantic information to optimize the exposure bias problem. After the pre-training model of large-scale Chinese corpus, the results are improved in LCSTS data set after the task tuning of the downstream summary, and the validity and stability of the model are verified.

S94 objective recognition not only classifies what category it is, but also detects sub-categories information to confirm the objective is what we want to find. Due to the problem that the number of labeled specific objective images is insufficient, and the subtle different between the specific objective and other objectives in the same category, we proposed an efficient method based on convolutional neural network (CNN) transfer learning. We adopted VGG16 network structure to learn low level visual feature, and the fully-connected layers were redesigned. The method retrained the last fully-connected layer and soft-max classifier, and fine-tuned the model with backpropagation algorithm. The experimental results show that the method makes it possible to use CNN fine-tuning for specific objective recognition with only small scale images, and effectively improves the accuracy of image classification.

A 1064 nm, 1 mJ pulsed fiber MOPA module, housed in 16”x14”x2.5” package for application in a lunar and planetary in-situ surface dating instrument is demonstrated. The module is based on a three-stage MOPA with a 60 μm core tapered fiber terminal amplifier. The master oscillator and first two preamplifier stages, which generate 20 μJ pulses, are all contained on a 13”x11”x1” board. Several improvements to the electronic signal control were instrumental to the laser development, including bipolar drive of the phase modulator for SBS suppression, shaping of the seed pulse to compensate pulse steepening, and pulsed operation of the power amplifier pump to reduce spontaneous emission at low pulse repetition frequency. The packaged laser runs at a repetition rate of 10 kHz and generates 10 ns pulses at 1 mJ with a 40 GHz linewidth, an M 2 ~ 1.2 beam quality, and an 18 dB polarization extinction ratio. The modular design enables seven independent lasers to be stacked in a 20”x18”x16.25” enclosure, supporting a path towards a fiber laser based LARIMS for advanced materials characterization and chronological dating in harsh and remote environments.

Due to the limited space of the carrier, the multi-antenna system on the carrier is prone to electromagnetic interference. Most of the researchers used genetic algorithms to optimize the antenna layout. On this basis, this paper proposes a method of interactive simulation of MATLAB and electromagnetic simulation software CST. Firstly, the position information of the antenna and the carrier platform is initialized in MATLAB, and then the program of the genetic algorithm is run, and the layout scheme of each generation is automatically transmitted to the CST through the ActiveX. The CST solves the antenna coupling degree and automatically returns it to MATLAB as the fitness value of the scheme. After dozens of generations of genetic algorithms, a better layout scheme can be found. Compared with the previous optimization methods, the method of this paper does not need to build complex electromagnetic numerical calculation model, and it has strong operability and high degree of visualization. Therefore, it has certain reference for the layout optimization of multi-antenna systems of other carriers.

Aircraft detection is one of the applications of high resolution remote-sensing images. Conventional method of aircraft detection use deep and complicated network, which needs highly computing cost. In this letter, an efficient and effective aircraft detection framework based on BOVW features and cascade AdaBoost classifier has been proposed. A variety of affine invariant features, which represent the complicate structure of aircraft, are extracted from sliding windows, and the direction estimation of aircraft is introduced to align the aircraft before detection. Besides, an accurate NMS algorithm is designed to make the target location more accurate. The proposed method is evaluated with recent advanced detection methods. The superior experimental outcome indicates that our framework achieves better accuracy as well as efficiency.

Aiming at the problem of objects in image recognition rectangle, but objects which are input into convolutional neural networks square, the object recognition model was put forward which was based on BING method to realize object estimate, used vectorization of convolutional neural networks to realize input square image in convolutional networks, therefore, built joint convolution neural networks, which achieve multiple size image input. Verified by experiments, the accuracy of multi-object image recognition was improved by 6.70% compared with single vectorization of convolutional neural networks. Therefore, image recognition method of joint convolutional neural networks can enhance the accuracy in image recognition, especially for target in rectangular shape.

We demonstrate high power, deep ultraviolet (DUV) conversion to 266 nm through frequency quadrupling of a nanosecond pulse width 1064 nm fiber master oscillator power amplifier (MOPA). The MOPA system uses an Yb-doped double-clad polarization-maintaining large mode area tapered fiber as the final gain stage to generate 0.5-mJ, 10 W, 1.7- ns single mode pulses at a repetition rate of 20 kHz with measured spectral bandwidth of 10.6 GHz (40 pm), and beam qualities of Mx 2=1.07 and My 2=1.03, respectively. Using LBO and BBO crystals for the second-harmonic generation (SHG) and fourth-harmonic generation (FHG), we have achieved 375 μJ (7.5 W) and 92.5 μJ (1.85 W) at wavelengths of 532 nm and 266 nm, respectively. To the best of our knowledge these are the highest narrowband infrared, green and UV pulse energies obtained to date from a fully spliced fiber amplifier. We also demonstrate high efficiency SHG and FHG with walk-off compensated (WOC) crystal pairs and tightly focused pump beam. An SHG efficiency of 75%, FHG efficiency of 47%, and an overall efficiency of 35% from 1064 nm to 266 nm are obtained.

Duchenne muscular dystrophy (DMD) is a degenerative muscle disorder characterized by the lack of dystrophin expression at the sarcolemma of muscle fibers. In addition, DMD patients acquire osteopenia, fragility fractures, and scoliosis indicating that a deficiency in skeletal homeostasis coexists but little is known about the effects of DMD on bone and other connective tissues within the musculoskeletal system. Recent evidence has emerged implicating adult stem cell dysfunction in DMD myopathogenesis. Given the common mesenchymal origin of muscle and bone, we sought to investigate bone and other musculoskeletal tissues in a DMD mouse model. Here, we report that dystrophin-utrophin double knockout (dko) mice exhibit a spectrum of degenerative changes, outside skeletal muscle, in bone, articular cartilage, and intervertebral discs, in addition to reduced lifespan, muscle degeneration, spinal deformity, and cardiomyopathy previously reported. We also report these mice to have a reduced capacity for bone healing and exhibit spontaneous heterotopic ossification in the hind limb muscles. Therefore, we propose the dko mouse as a model for premature musculoskeletal aging and posit that a similar phenomenon may occur in patients with DMD.

In order to improve the accuracy of prediction for software defect data sets, competitive organization coevolutionary algorithm is presented and applied for dealing with the software defect data. During this algorithm, mechanism of competition is introduced into coevolutionary algorithm. Then leagues are formed based on the importance of attributes among them. And three evolution operators which are reduced operator, allied operators and disturbed operators are developed. Furthermore, both the importance of attributes and evaluation from competition are used for the calculation of fitness function. Finally, five data sets from NASA MDP(Metrics Data Program) were used to validate the algorithm. The experimental results show that the proposed algorithm is effective for software defect prediction.

Fibrosis is the result of an excessive amount of fibrous connective tissue deposited into the extracellular matrix (ECM) space of damaged tissues from injury or disease. Collagens, particularly types I and III are the main constituents of the fibrotic scar tissue as well as a mixture of fibrotic cells. Severely fibrotic tissue will develop chronic healing problems resulting in tissue/organ dysfunction. More attention needs to be given to the fibrotic differentiation and related effects in bioengineered tissues. The current review provides an update on the mechanism behind fibrosis formation as well as technical measurements and preventions.

We review the progress made by us on the exploration of the fundamental limits to terahertz (THz) generation from several semiconductor electrooptic materials. Through the measurements of the THz output versus the pump beam in terms of incident angle, polarization, azithumal angle, and pump intensity, we have demonstrated that we can precisely determine the contributions made by the optical rectification and photocurrent surge. When a material is pumped below its bandgap, optical rectification is always the mechanism for the THz generation. Above the bandgap, however, these two mechanisms often compete with each other, depending on the material characteristics and pump intensity. At a sufficiently high pump intensity, optical rectification usually becomes the dominant mechanism for a second-order nonlinear material. Our analysis indicates that second-order nonlinear coefficients are resonantly enhanced when a material is pumped above its bandgap. In such a case, the THz output power and normalized conversion efficiency can be dramatically increased. We have also illustrated that, for some materials, two-photon absorption can be one of the fundamental limits to the THz generation.

We have investigated power scaling for the efficient generation of the broadband terahertz (THz) pulses. These THz short pulses are converted from ultrafast laser pulses propagating in a class of semiconductor electrooptic materials. By measuring the dependence of the THz output on the pump beam in terms of incident angle, polarization, azimuthal angle, and pump intensity, we have precisely determined the contributions made by optical rectification, drift of carriers under a surface or external field, and photo-Dember effect. When a second-order nonlinear material is pumped below its bandgap, optical rectification is always the mechanism for the THz generation. Above the bandgap, however, the three mechanisms mentioned earlier often compete with one another, depending on the material characteristics and pump intensity. At a sufficiently high pump intensity, optical rectification usually becomes the dominant mechanism for a second-order nonlinear material. Our analysis indicates that second-order nonlinear coefficients can be resonantly enhanced when a material is pumped above its bandgap. In such a case, the THz output power and normalized conversion efficiency can be dramatically increased. We have also analyzed how the THz generation is affected by some competing processes such as two-photon absorption.

Although we recognize the many advantages of improved musculoskeletal health, we also note that our ability to sustain this health and to maintain quality of life in an aging population is currently deficient. However, global efforts have produced numerous advances in tissue engineering and regenerative medicine that will collectively serve to fill this deficiency in the near future. The purpose of this review is to highlight our current knowledge, to outline our recent advances, and to discuss the evolving paradigms in skeletal muscle injury and repair.

Skeletal muscle atrophy in cancer cachexia is mediated by the interaction between muscle stem cells and various tumor factors. Although Notch signaling has been known as a key regulator of both cancer development and muscle stem cell activity, the potential involvement of Notch signaling in cancer cachexia and concomitant muscle atrophy has yet to be elucidated. The murine K7M2 osteosarcoma cell line was used to generate an orthotopic model of sarcoma-associated cachexia, and the role of Notch signaling was evaluated. Skeletal muscle atrophy was observed in the sarcoma-bearing mice, and Notch signaling was highly active in both tumor tissues and the atrophic skeletal muscles. Systemic inhibition of Notch signaling reduced muscle atrophy.In vitrococulture of osteosarcoma cells with muscle-derived stem cells (MDSCs) isolated from normal mice resulted in decreased myogenic potential of MDSCs, while the application of Notch inhibitor was able to rescue this repressed myogenic potential. We further observed that Notch-activating factors reside in the exosomes of osteosarcoma cells, which activate Notch signaling in MDSCs and subsequently repress myogenesis. Our results revealed that signaling between tumor and muscle via the Notch pathway may play an important role in mediating the skeletal muscle atrophy seen in cancer cachexia.

We have demonstrated mJ pulse energy with nanosecond pulses, narrow linewidth, and single-mode beam quality operation in a master oscillator power amplifier (MOPA) fiber laser system using an Yb-doped polarization maintaining tapered amplifier.

We have reviewed mechanisms for the efficient conversion of the broadband terahertz (THz) pulses from ultrafast laser pulses propagating in semiconductor electro-optic materials. By measuring the behaviors of the THz output versus the pump beam in terms of incident angle, polarization, and azimuthal angle, we have demonstrated that we can precisely determine the contributions made by optical rectification and photocurrent surge. When a material is pumped below its bandgap, optical rectification is always the mechanism for the THz generation. Above the bandgap, however, these two mechanisms often compete with each other, depending on the material characteristics and pump intensity. At a sufficiently high pump intensity, optical rectification usually becomes the dominant mechanism for a second-order nonlinear material. Our analysis indicates that second-order nonlinear coefficients can be resonantly enhanced when a material is pumped above its bandgap. In such a case, the THz output power and normalized conversion efficiency can be greatly enhanced. We have also analyzed how the THz generation is affected by some competing processes such as two-photon absorption.

We review our recent results on the investigation of carrier dynamics of semiconductor nanostructures, i.e., dot-in-a-well. We introduced a technique based on the measurement of time-resolved differential photoluminescence spectra induced by subpicosecond pump and probe laser pulses. In this technique, the temporal resolution was achieved by adjusting the temporal delay between the pump and probe pulses. Using such a new technique, we measured the exciton decay times and then studied the dependences of integrated photoluminescence intensity and photoluminescence linewidth on temperature in InAs quantum dots embedded in InGaAs/GaAs quantum wells. We gave consistent interpretations to our experimental results.

This article puts forward a kind of huge amounts of multi-object image recognition method -- BVCNN. Firstly, BING method is used to recognize images, which greatly reduces the time of estimating image targets, and makes it possible that quickly identify multiple target images, compared to traditional convolution neural networks only achieving single target image recognition, Secondly, vectorization of deep convolutional neural networks is used for deep learning of characteristics in local image and recognition, which speeds up network training and testing, thirdly, using the context information in multi-object image classification, to a certain extent, helps to distinguish individual of similar characteristics according to environment, improving the multi-object image recognition accuracy. According to experiments, identifying a single image by this model only need less than 1 s, and this model can be used for image information fusion.

Single-clad and double-clad Yb:YAG crystalline fiber waveguides (CFWs) have been prepared with Adhesive-Free Bonding (AFB®) technology. By using a fiber coupled laser diode as pump source, a single-mode laser with near diffraction limited beam quality M 2 =1.02 has been demonstrated in a double-clad CFW. The laser output power and efficiency are 13.2 W and 34%, respectively. In a single-clad CFW, core pumping was used. The laser output has top-hat beam profile. An output power of 28 W and a slope efficiency of 78% have been achieved respectively. Keywords: Crystalline fiber, waveguide, laser, master oscillator power amplifier, adhesive-free bond 1. INTRODUCTION There is great interest in making rare-earth (RE) doped YAG crystalline fiber lasers due to their higher power scalability compared to glass fibers [1]. However, current fiber growth (or pulling) techniques, such as laser heated pedestal growth (LHPG) [2-4] and micro pulling down [5], have not yet been successful in producing low loss, double-clad, or even single-clad fully crystalline fibers. As an alternative to conventional fiber pulling technologies, Onyx Optics’ well established Adhesive-Free Bond (AFB®) technology circumvents the problems with the fabrication of low-loss crystalline fiber waveguides (CFW) [6-8]. In the present work , we report on the first laser diode (LD) cladding pumped single-mode 1030-nm laser in a double-clad Yb:YAG CF W with a near diffraction-limited beam quality of M

A new solution for building high power, solid state lasers for space flight is to fabricate the whole laser resonator in a single (monolithic) structure or alternatively to build a contiguous diffusion bonded or welded structure. Monolithic lasers provide numerous advantages for space flight solid-state lasers by minimizing misalignment concerns. The closed cavity is immune to contamination. The number of components is minimized thus increasing reliability. Bragg mirrors serve as the high reflector and output coupler thus minimizing optical coatings and coating damage. The Bragg mirrors also provide spectral and spatial mode selection for high fidelity. The monolithic structure allows short cavities resulting in short pulses. Passive saturable absorber Q-switches provide a soft aperture for spatial mode filtering and improved pointing stability. We will review our recent commercial and in-house developments toward fully monolithic solid-state lasers.

Among almost all the work done on parametric processes in the past, two-photon absorption (TPA) has been intentionally avoided by properly choosing the pump wavelength. On the contrary, we have purposely explored how TPA sets the fundamental limit on the optical parametric generation (OPG) and oscillation (OPO) by using periodically poled LiNbO 3 (PPLN) as a testing nonlinear medium. Indeed, we have demonstrated that the TPA dramatically affects the characteristics of these two parametric processes. To our surprise, the OPO is actually not a preferred parametric process anymore compared with the OPG due to the presence of the strong TPA. In addition, the TPA can further be enhanced by self-focusing effect caused by accumulated thermal effect (ATE).

A frequency converter is designed for producing efficient blue light at 0.4396 μm from a monochromatic, pulsed, 1.3188-μm Nd-doped emission line by quasi-phase-matched (QPM) second-harmonic generation (SHG), followed by sum-frequency generation, in an integrated, periodically poled KTiOPO4 (KTP) and monodomain KTP crystal. An average 55-mW (peak power of 43.0 W) blue light is achieved by us for an average 2.5-W (peak power of 2.0 kW) pump power; the overall conversion efficiency is thus 2.2%. Through polarization optimization, the same device produces an average 435 mW (peak power of 339.8 W) for the red light through QPM SHG, with a corresponding conversion efficiency of 16%. As a practical blue-light converter, the angle- and temperature-tuning characteristics and period tolerance have been fully investigated.

The prognosis of colorectal carcinoma (CRC) is unstable in the stage II-III patients. Patients with early stage II CRC have a relative poor prognosis while other stage III CRC patients have a better prognosis. In our work, by utilizing the expression of lncRNAs and mRNAs measured by microarray (GSE39582), we constructed a risk score staging system with Cox multivariate regression model to predict the outcome of grade II-III CRC patients. Ten genes including two lncRNAs and eight mRNAs were used to estimate the survival of stage II-III CRC patients. The patients with high risk scores have poorer survival rate those with low risk scores, significantly. These results were further validated in another three independent datasets (GSE37892, GSE33113, and GSE17536). The relationship between clinical information and were evaluated, and the risk score is independent from the other clinical information and performs better in evaluating the survival of stage II-III CRC patients. Moreover, the correlation between chemotherapy was also evaluated, and we found that both patients with or without chemotherapy have a poor survival in high risk group. Gene Set Enrichment Analysis were used to find the difference between high-risk and low-risk groups, and pathways including cell adhesion and focal adhesion were significantly enriched, suggesting that the risk score reflects the status of cell-cell physical interaction. In summary, we constructed a risk staging model for grade II-III CRC, which is independent from and performs better than clinical information.

OFDM-CPM is now a recognized alternative to conventional OFDM for the transmission of signals over Non-Continuous Wideband channel. This paper presents a characterization of nonlinear distortion effects in orthogonal frequency-division multiplexing/continuous phase modulation (OFDM-CPM) transmission systems on Non-Continuous Wideband. In a second step, a comparison with simulation results is carried out to show that OFDM-CPM is more suitable to attain high spectrum efficiency than OFDM-QPSK.

Double-clade crystalline fiber waveguide (CFW) has been produced by using adhesive-free bond (AFB®) technology. The waveguide consists of a 1 at.% Yb:YAG core, un-doped YAG inner cladding and ceramic spinel outer cladding. It is a direct analog of the conventional double-clad glass fiber laser in the crystal domain. Signal gain of 45 or 16.5 dB has been measured in a preliminary master oscillator power amplifier (MOPA) experiment. Due to the high laser gain and the weak Fresnel reflection at the uncoated waveguide ends, the CFW even starts self-lasing above a certain pump power. Laser output power of 4 W in the backward propagation direction has been measured for input pump power of 44 W. After considering the same amount of forward propagated laser power, the laser efficiency to the absorbed pump power is estimated to be about 44%. In principle, CFW can have extremely large single mode area for high efficiency and high power laser applications. So far, Single mode area < 6700 μm2 has been demonstrated in Er:YAG CFWs.

Osteosarcoma (OS), chondrosarcoma (CSA), and Ewings sarcoma (ES) are the most common primary malignancies of bone, and are rare diseases. As with all sarcomas, the prognosis of these diseases ultimately depends on the presence of metastatic disease. Survival is therefore closely linked with the biology and metastatic potential of a particular bone tumor’s cells. Here we describe a significant correlation of aldehyde dehydrogenase (ALDH) activity and the presence/absence of distant metastases in ten consecutive cases of human bone sarcomas. Additionally, cultured human CSA cells, which are historically chemo- and radio-resistant, may be sensitive to the ALDH inhibitor, disulfiram. While it is premature to draw broad conclusions from such a small series, the importance of ALDH activity and inhibition in the metastatic potential of primary bone sarcomas should be investigated further.

Referring to the problem of SAR image registration, an image registration method based on Scale Invariant Feature Transform (SIFT) and Multi-Scale Autoconvolution (MSA) is proposed. Based on the extraction of SIFT descriptors and the MSA affine invariant moments of the region around the keypoints, the feature fusion method based on canonical correlation analysis (CCA) is employed to fuse them together to be a new descriptor. After the control points are rough matched, the distance and gray correlation around the rough matched points are combined to build the similarity matrix and the singular value decomposition (SVD) method is employed to realize precise image registration. Finally, the affine transformation parameters are obtained and the images are registered. Experimental results show that the proposed method outperforms the SIFT method and achieves high accuracy in sub-pixel level.

This paper presents the experimental results of two-wave mixing and self-pumped phase conjugation (SPPC) in barium titanate (BaTiO3) crystals doped with Rh or Ce. It has been shown that the main parameters of SPPC depend very strongly on the concentrations of dopants, the incident angles and the crystal cut used. High gain in two-wave mixing corresponds to low phase-conjugate reflectivity in all samples. It is impossible to maintain the highest reflectivity and gain simultaneously, because the gain will definitely be limited when we try to get the highest reflectivity, and vice versa. Some qualitative explanation of the experimental results is also given.

For the first time, we demonstrated backward second-harmonic generation (SHG) in a submicron-period ion-exchanged KTiOPO 4 (KTP) waveguide. Such a period is the shortest ever achieved in ferroelectric materials. The characteristics of backward quasi-phase-matching at the 6th and 7th orders were investigated. Our analyses on the backward SHG show that the second-harmonic linewidth and conversion efficiency are very sensitive to the grating period error. According to our measurement on the backward SH linewidth, we determined the period error to be ∼4.5×10 −3 μm for our sample, which is close to the value of ∼7×10 −3 μm measured by other techniques.

We review our recent results on efficient generation of coherent blue/green light based on different configurations and structures. We have used second-harmonic generation with forward or backward configuration in short-period periodically-poled bulk and wave-guide KTP to generate blue light. We have used backward second-harmonic generation to characterize periodically-segmented submicron KTP waveguides. By cascading second-harmonic generation and subsequent sum-frequency generation, we have generated coherent blue light in KTP and/or Ce:KTP crystals. We have also studied damage threshold of bulk KTP; we hope that we can eventually use it towards efficient generation of cw coherent green light.

The gain of forward two-wave mixing (FTWM) and time response of a Cu-doped barium strontium potassium sodium niobate (KNSBN) crystal are investigated at 514.5 nm in the temperature range 20–120°C. It is found that the gain reaches its maximum value at about 90°C. The effective photorefractive charge density decreases with temperature. The response and erasure rates increase with increasing temperature. Moreover, the erasure process can be described by a diexponential function. A band diagram of the two-center charge transport model is presented. The phenomena are explained qualitatively.