Face recognition is a hot research field in computer vision, and it has high practical value for the detection of specific sensitive characters. Research found that traditional hand-crafted features, there are uncontrolled environments such as pose, facial expression, illumination occlusion influencing accuracy poor performance, so deep learning method adopted. On basis face detection, Convolutional Neural Network (CNN) based on TensorFlow, an open source framework, proposed recognition....

Face recognition is a kind of biometrics which based on the facial feature information human. And face has wide application value in computer security, medical treatment, security monitoring, human-computer interaction and finance. extraction key technology, it related to selection algorithm. Local Binary Pattern texture description method that describes local an image gray-scale range. In recent years, many researchers have successfully applied recognition, achieved remarkable results....

Permeability and formation factor are physical properties of porous rocks useful for assessing reservoirs. Neither property varies consistently as porosity varies. The relationship both to is complex, being sensitive the structure microstructure, i.e., sizes pore throats, numbers pores, relationships between pores throats. Physical models account these factors require parameters that describe physically relevant microstructure. A partial characterization throats embodied in type throat size....

Recently, few certified defense methods have been developed to provably guarantee the robustness of a text classifier adversarial synonym substitutions. However, all existing assume that defenders are informed how adversaries generate synonyms, which is not realistic scenario. In this paper, we propose certifiably robust method by randomly masking certain proportion words in an input text, above unrealistic assumption no longer necessary. The proposed can defend against only word...

Deep neural networks are vulnerable to adversarial attacks, where a small perturbation an input alters the model prediction. In many cases, malicious inputs intentionally crafted for one can fool another model. this paper, we present first study systematically investigate transferability of examples text classification models and explore how various factors, including network architecture, tokenization scheme, word embedding, capacity, affect examples. Based on these studies, propose genetic...

Purpose Management response to consumer comments has become a widely adopted marketing strategy address the undesirable effects caused by negative remarks. Yet, when and what management is more effective under circumstances remains under-researched. This study aims fill this gap. Design/methodology/approach In three experiments using five different products, authors manipulate psychological construal level (psychological distance: distant vs proximal) (response of primary secondary features)...

Starburst algorithm that combines feature-based and model-based approaches, can achieve a good tradeoff between run-time performance accuracy for dark-pupil infrared illumination. On this basis, eye images are preprocessed using AdaBoost classifier to extract region of interest which contains pupil in paper. Thus, we ensure the start point model is located within region, reducing number iterations algorithm. Meanwhile, reflection points detected filled effectively, eliminating interference...

Single chaotic system which is realized with limited precisions in digital computers can lead to dynamical degradation of the original systems. To solve problem, this paper introduces a new kind dual encryption algorithm constructed via Logistic map and Sine iterative map. Digital PR sequence generated using system, then an E-mail based on implemented by VC++ language. In comparison single expend control parameters increase complexity chaos behavior so as improve security encryption. order...

The research of biomedical data is crucial for disease diagnosis, health management, and medicine development. However, are usually characterized by high dimensionality class imbalance, which increase computational cost affect the classification performance minority class, making accurate difficult. In this paper, we propose a method based on feature selection resampling. First, use minimal-redundancy maximal-relevance (mRMR) to select features, reduce dimension, cost, improve generalization...

Dietary crude protein concentration oscillation can improve the nitrogen utilization efficiency of ruminants. However, little is known about dynamic changes in microbiota and fermentation rumen calves during period. In this study, six were fed an oscillating diet at 2-day intervals, including a high-protein (HP) low-protein (LP). The parameters, plasma urea-N concentration, bacterial diversity characterized throughout concentrations volatile fatty acids, NH

Summary In seismic prospecting, the first-break travelling time has been widely used. As industry moves toward denser source and receiver coverage, high-density data which leads to a dramatic increase in fold number, workload of picking is getting larger larger. Many techniques applied picking, many traditional are based on energy ratios. Here we proposed method mathematical morphological top-hat transformation achieve picking. Then use synthetic field examples shown performance this method....

In order to improve performance and robustness of clustering, it is proposed generate aggregate a number primary clusters via clustering ensemble technique. Fuzzy approaches attempt the fuzzy tasks. However, in these approaches, cluster (or clustering) reliability has not paid much attention to. Ignoring makes weak dealing with low-quality base methods. this paper, we have utilized unreliability estimation local weighting strategy propose new method which introduced Reliability Based...

It is still a challenging task to learn neural text generation model under the framework of generative adversarial networks (GANs) since entire training process not differentiable. The existing strategies either suffer from unreliable gradient estimations or imprecise sentence representations. Inspired by principle sparse coding, we propose SparseGAN that generates semantic-interpretable, but representations as inputs discriminator. key idea treat an embedding matrix over-complete...