Since it can effectively address the problem of sparsity and cold start collaborative filtering, knowledge graph (KG) is widely studied employed as side information in field recommender systems. However, most existing KG-based recommendation methods mainly focus on how to encode associations KG, without highlighting crucial signals which are latent user-item interactions. As such, learned embeddings underutilize two kinds pivotal insufficient represent semantics users items vector space.

Location-based service (LBS) has been widely used in various fields of industry, and become a vital part people's daily life. However, while providing great convenience for users, LBS results serious threat on users' location privacy, due to its more untrusted server-side. In this article, we propose privacy-preserving system by constructing "cover-up ranges" protect the query ranges associated with sequence. Firstly, present client-based framework privacy protection LBS, which requires no...

CAPTCHA is now a standard security technology for differentiating between computers and humans, the most widely deployed schemes are text-based. While many text have been broken, hollow CAPTCHAs emerged as one of latest designs, they by major companies such Yahoo!, Tencent, Sina, China Mobile Baidu. A main feature to use contour lines form connected characters with aim improving usability simultaneously, it hard techniques segment recognize characters, which however easy human eyes. In this...

Knowledge graph completion (KGC) is the task of predicting missing links based on known triples for knowledge graphs. Several recent works suggest that Graph Neural Networks (GNN) exploit structures achieve promising performance KGC. These models learn information called messages from neighboring entities and relations then aggregate to update central entity representations. The drawback existing GNN lies in they tend treat equally fixed network parameters, overlooking distinction each...

Landmark codes underpin reliable physical layer communication, e.g., Reed-Muller, BCH, Convolution, Turbo, LDPC and Polar codes: each is a linear code represents mathematical breakthrough. The impact on humanity huge: of these has been used in global wireless communication standards (satellite, WiFi, cellular). Reliability over the classical additive white Gaussian noise (AWGN) channel enables benchmarking ranking different codes. In this paper, we construct KO codes, computationaly...

Identifying objects of interest from digital vision signals is a core task intelligent systems. However, fast and accurate identification small moving targets in real-time has become bottleneck the field target detection. In this paper, problem detection fast-moving printed circuit board (PCB) tiny investigated. This very challenging because PCB defects are usually compared to whole board, due pursuit production efficiency, actual speed fast, which puts higher requirements on To end, new...

Unsupervised image Anomaly Detection (UAD) aims to learn robust and discriminative representations of normal samples. While separate solutions per class endow expensive computation limited generalizability, this paper focuses on building a unified framework for multiple classes. Under such challenging setting, popular reconstruction-based networks with continuous latent representation assumption always suffer from the "identical shortcut" issue, where both abnormal samples can be well...

Reed-Muller (RM) codes are conjectured to achieve the capacity of any binary-input memoryless symmetric (BMS) channel, and observed have a comparable performance that random in terms scaling laws. On negative side, RM lack efficient decoders with close maximum likelihood decoder for general parameters. Also, they only admit certain discrete sets rates. In this paper, we focus on subcodes flexible rates can take code dimension from 1 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML"...

FaceDCAPTCHA and FR-CAPTCHA, proposed in 2014, are both face-based CAPTCHAs relying on human face recognition. The security of is based the difficulty classifying real faces fake while FR-CAPTCHA finds two belonging to same person a complex background. In this paper, edge detection employed obtain small then an SVM classifier used differentiate images with color texture, LBP, SIFT Laws' Masks features extracted from faces. attack success rate 48%. OpenCV utilized detect four find most...

Generative machine learning models are being increasingly viewed as a way to share sensitive data between institutions. While there has been work on developing differentially private generative modeling approaches, these approaches generally lead sub-par sample quality, limiting their use in real world applications. Another line of focused which higher quality samples but currently lack any formal privacy guarantees. In this work, we propose the first framework for membership estimation...

In statistical learning and analysis from shared data, which is increasingly widely adopted in platforms such as federated meta-learning, there are two major concerns: privacy robustness. Each participating individual should be able to contribute without the fear of leaking one's sensitive information. At same time, system robust presence malicious participants inserting corrupted data. Recent algorithmic advances data focus on either one these threats, leaving vulnerable other. We bridge...

Information abounds in all fields of the real life, which is often recorded as digital data computer systems and treated a kind increasingly important resource. Its increasing volume growth causes great difficulties both storage analysis. The massive cloud environments has significant impacts on quality service (QoS) systems, becoming an challenging problem. In this paper, we propose multiobjective optimization model for reliable clouds through considering cost reliability simultaneously....

The crowd counting task aims at estimating the number of people located in an image or a frame from videos. Existing methods widely adopt density maps as training targets to optimize point-to-point loss. While testing phase, we only focus on differences between numbers and global summation maps, which indicate inconsistency evaluation criteria. To solve this problem, introduce new target, named local map (LCM), obtain more accurate results than based approaches. Moreover, also propose...

Differential privacy has become a widely accepted notion of privacy, leading to the introduction and deployment numerous privatization mechanisms. However, ensuring guarantee is an error-prone process, both in designing mechanisms implementing those Both types errors will be greatly reduced, if we have data-driven approach verify guarantees, from black-box access mechanism. We pose it as property estimation problem, study fundamental trade-offs involved accuracy estimated guarantees number...

Transductive Few-Shot Learning (TFSL) has recently attracted increasing attention since it typically outperforms its inductive peer by leveraging statistics of query samples. However, previous TFSL methods usually encode uniform prior that all the classes within samples are equally likely, which is biased in imbalanced and causes severe performance degradation. Given this pivotal issue, work, we propose a novel Conditional Transport (CT) based model called {\textbf P}rototypes-oriented...

Scientific WorkFlows (SWFs) play significant roles in scientific research and engineering simulation, which are often data intensive has complex dependencies. The storage of massive intermediate datasets great impacts on the performance quality service a SWF system, become difficult task. Through analyzing cost transitive tournament shortest path (CTT-SP)-based algorithm proposed for systems, disadvantage CTT-SP over its sensitivity to main branches can be found. We an improved based...

We study the canonical statistical task of computing principal component from $n$ i.i.d.~data in $d$ dimensions under $(\varepsilon,\delta)$-differential privacy. Although extensively studied literature, existing solutions fall short on two key aspects: ($i$) even for Gaussian data, private algorithms require number samples to scale super-linearly with $d$, i.e., $n=\Omega(d^{3/2})$, obtain non-trivial results while non-private PCA requires only $n=O(d)$, and ($ii$) techniques suffer a...

We introduce a universal framework for characterizing the statistical efficiency of estimation problem with differential privacy guarantees. Our framework, which we call High-dimensional Propose-Test-Release (HPTR), builds upon three crucial components: exponential mechanism, robust statistics, and mechanism. Gluing all these together is concept resilience, central to estimation. Resilience guides design algorithm, sensitivity analysis, success probability analysis test step in...