Background: The side effects of drugs are not only harmful to humans but also the major reasons for withdrawing approved drugs, bringing greater risks pharmaceutical companies. However, detecting a given drug via traditional experiments is time- consuming and expensive. In recent years, several computational methods have been proposed predict drugs. most cannot effectively integrate heterogeneous properties Methods: this study, we adopted network embedding method, Mashup, extract essential...

Drugs are an important way to treat various diseases. However, they inevitably produce side effects, bringing great risks human bodies and pharmaceutical companies. How predict the effects of drugs has become one essential problems in drug research. Designing efficient computational methods is alternative way. Some studies paired effect as a sample, thereby modeling problem binary classification problem. selection negative samples key this case. In study, novel sample strategy was designed...

While widely adopted in practical applications, face recognition has been critically discussed regarding the malicious use of images and potential privacy problems, e.g., deceiving payment system causing personal sabotage. Online photo sharing services unintentionally act as main repository for crawler applications. This work aims to develop a privacy-preserving solution, called Adversarial Privacy-preserving Filter (APF), protect online shared from being maliciously used.We propose an...

Turing test was originally proposed to examine whether machine's behavior is indistinguishable from a human. The most popular and practical CAPTCHA, which discriminate algorithm human by offering recognition-alike questions. recent development of deep learning has significantly advanced the capability in solving CAPTCHA questions, forcing designers increase question complexity. Instead designing questions difficult for both human, this study attempts employ limitations design robust easily...

Attributed to the unstructured and sparse nature of point clouds, transformer shows greater potential in clouds data processing. However, recent query-based 3D detectors usually project features acquired from a backbone into structured compact Bird's Eye View(BEV) plane before adopting transformer, which destroys sparsity features, introducing empty tokens additional resource consumption for transformer. To this end, paper, we propose novel detector called Clusterformer, our Clusterformer...

Low rank matrix algorithm has attracted widely attention since it was put forward. There are many algorithms proposed to improve it. Most methods take each sample as a column, but Robust Generalized Rank Approximations of Matrices(RGLRAM) treats matrix, thus we can find the low approximations on collection matrices not just single matrix. RGLRAM only robust small Gaussian noise also large sparse noise. In these cases, extend video approximation. The be treated images. Each frame is divided...

Drugs are very important for human life because they can provide treatment, cure, prevention, or diagnosis of different diseases. However, also cause side effects, which increase the risks humans and pharmaceuticals companies. It is essential to identify drug effects in discovery. To date, lots computational methods have been proposed predict drugs most them used fact that similar always effects. previous studies did not analyze substructures highly related kind effect.In this study, we...

3D object detection has achieved great progress recently. However, the contradiction between high accuracy and rapid inference is a crucial issue, which particularly evident in voxel-based networks pillar-based networks. Voxel-based can achieve accuracy, but sparse convolution backbone blocks real-time model deployment. Pillar-based are deploymentfriendly inference, they cannot perform as accurately To reduce gap these two types of well keep speed networks, this paper, we propose Learn from...

In spite of the successful application in many fields, machine learning models today suffer from notorious problems like vulnerability to adversarial examples. Beyond falling into cat-and-mouse game between attack and defense, this paper provides alternative perspective consider example explore whether we can exploit it benign applications. We first attribute human-model disparity on employing non-semantic features. While largely ignored classical mechanisms, feature enjoys three interesting...

Abstract:Mn salts[MnSO4]-doped TiO2 electrodes thin film was synthesized by the hydrothermal method. To prepare working electrode, The or Mn-doped slurry coated onto fluorine-doped tin oxide glass substrate doctor blade method and then sintered at 450 °C. X-ray photoelectron spectroscopy (XPS) data indicated that doped Mn ions exist in forms of Mn7+. This study show a photovoltaic efficiency 5.15%, which is higher than undoped (4.14%) increase photovoltage fill factors 13.2% from 455 mV to...

When facing complex requirements of city development, ever-growing spatial data, rapid development geographical business and increasing complexity, collaboration between multiple users departments is needed urgently, however conventional GIS software (such as Client/Server model or Browser/Server model) are not support this well. Collaborative application one the good resolutions. has four main problems to resolve: consistency co-edit conflict, real-time responsiveness, unconstrained...

This tutorial discusses the trustworthiness issue in multimedia analysis. Starting from introducing two types of spurious correlations learned distilling human knowledge, we partition (visual) feature space along dimensions task-relevance and semantic-orientation. Trustworthy analysis ideally relies on task-relevant semantic features consists three modules as trainer, interpreter tester. These essentially form a closed loop, which respectively address goals extracting features, detecting to...

With the development of remote sensing technology, getting image data more and more, how to apply manage mass safely efficiently has become an urgent problem be solved. According methods characteristics management application, this paper puts forward a new method that takes Oracle Call Interface InterMedia store data, then component realize system function modules. Finally, it successfully VC storage management.

Analytical structure of fuzzy controller has recovered two important issues: why is useful in application and what relationships are there between the most used PID controller. But these existing outcomes uniformly studied under conditions that all rules taken same membership degrees. This not real some applications since have likely different certainties or fuzziness. In this paper we study analytical case take degrees, graphical comparison with control system. The deriving generalize...

Deep learning models suffer from the problem of semantic discontinuity: small perturbations in input space tend to cause semantic-level interference model output. We argue that discontinuity results these inappropriate training targets and contributes notorious issues such as adversarial robustness, interpretability, etc. first conduct data analysis provide evidence existing deep models, then design a simple continuity constraint which theoretically enables obtain smooth gradients learn...

While widely adopted in practical applications, face recognition has been disputed on the malicious use of images and potential privacy issues. Online photo sharing services accidentally act as main approach for crawlers to exploit access portrait privacy. In this demo, we propose an adversarial privacy-preserving filter, which can preserve image from algorithms. This filter is generated by end-cloud collaborated attack framework consisting three modules: (1) Image-specific gradient...