Deep Random Walk (DeepWalk) can learn a latent space representation for describing the topological structure of network.However, relational network classification, DeepWalk be suboptimal as it lacks mechanism to optimize objective target task.In this paper, we present Discriminative (DDRW), novel method classification.By solving joint optimization problem, DDRW representations that well capture and meanwhile are discriminative classification task.Our experimental results on several real...

Network representation learning (NRL) aims to learn low-dimensional vectors for vertices in a network. Most existing NRL methods focus on representations from local context of (such as their neighbors). Nevertheless, many complex networks also exhibit significant global patterns widely known communities. It's intuitive that the same community tend connect densely and share common attributes. These are expected improve benefit relevant evaluation tasks, such link prediction vertex...

Session-based target behavior prediction aims to predict the next item be interacted with specific types (e.g., clicking). Although existing methods for session-based leverage powerful representation learning approaches encode items’ sequential relevance in a low-dimensional space, they suffer from several limitations. Firstly, focus on only utilizing same type of user prediction, but ignore potential taking other data as auxiliary information. This is particularly crucial when sparse...

Cross-domain recommendation (CDR) aims to provide better results in the target domain with help of source domain, which is widely used and explored real-world systems. However, CDR matching (i.e., candidate generation) module struggles data sparsity popularity bias issues both representation learning knowledge transfer. In this work, we propose a novel Contrastive Cross-Domain Recommendation (CCDR) framework for matching. Specifically, build huge diversified preference network capture...

Real-world integrated personalized recommendation systems usually deal with millions of heterogeneous items. It is extremely challenging to conduct full corpus retrieval complicated models due the tremendous computation costs. Hence, most large-scale consist two modules: a multi-channel matching module efficiently retrieve small subset candidates, and ranking for precise recommendation. However, suffers from cold-start problems when adding new channels or data sources. To solve this issue,...

Adversarial patch is an important form of real-world adversarial attack that brings serious risks to the robustness deep neural networks. Previous methods generate patches by either optimizing their perturbation values while fixing pasting position or manipulating patch's content. This reveals positions and perturbations are both attack. For that, in this article, we propose a novel method simultaneously optimize for patch, thus obtain high success rate black-box setting. Technically, regard...

Log anomaly detection is a key component in the field of artificial intelligence for IT operations (AIOps). Considering log data variant domains, retraining whole network unknown domains inefficient real industrial scenarios. However, previous deep models merely focused on extracting semantics sequences same domain, leading to poor generalization multi-domain logs. To alleviate this issue, we propose unified Transformer-based framework (LogFormer) improve ability across different where...

Sequential recommendation and group are two important branches in the field of recommender system. While considerable efforts have been devoted to these an independent way, we combine them by proposing novel sequential problem which enables modeling dynamic representations is crucial for achieving better performance. The major challenge how effectively learn based on user-item interactions members past time frames. To address this, devise a Group-aware Long- Short-term Graph Representation...

In tag-enhanced video recommendation systems, videos are attached with some tags that highlight the contents of from different aspects. Tag ranking in such systems provides personalized tag lists for their candidates. A better model could attract users to click more tags, enter corresponding channels, and watch tag-specific videos, which improves both rate watching time. However, most conventional models merely concentrate on tag-video relevance or tag-related behaviors, ignoring rich...

Real-world recommendation systems need to deal with millions of item candidates. Therefore, most practical large-scale usually contain two modules. The matching module aims efficiently retrieve hundreds high-quality items from large corpora, while the ranking generate specific ranks for these items. Recommendation diversity is an essential factor that strongly impacts user experience. There are lots efforts have explored in ranking, should take more responsibility diversity. In this article,...

Click-through rate (CTR) prediction plays a critical role in recommender systems and other applications. Recently, modeling user behavior sequences attracts much attention brings great improvements the CTR field. Many existing works utilize mechanism or recurrent neural networks to exploit interest from sequence, but fail recognize simple truth that user's real-time interests are inherently diverse fluid. In this paper, we propose DisenCTR, novel dynamic graph-based disentangled...

With the continuous advancement of machine learning, numerous malware detection methods that leverage this technology have emerged, presenting new challenges to generation adversarial malware. Existing function-preserving attacks fall short effectively modifying portable executable (PE) control flow graphs (CFGs), thereby failing bypass graph neural network (GNN) models utilize CFGs for detection. To solve issue, we introduce a novel base modification method called active opcode insertion,...

Capturing users' precise preferences is of great importance in various recommender systems (eg., e-commerce platforms), which the basis how to present personalized interesting product lists individual users. In spite significant progress has been made consider relations between users and items, most existing recommendation techniques solely focus on singular type user-item interactions. However, interactive behavior often exhibited with multi-type (e.g., page view, add-to-favorite purchase)...

Session-based recommendation aims at predicting the next item that a user is more likely to interact with by target behavior type. Most of existing session-based methods focus on developing powerful representation learning approaches model items' sequential correlations, whereas they usually encounter following limitations. Firstly, only utilize sessions belong type, neglecting potential leveraging other types as auxiliary information for modeling preference. Secondly, separately...

This article investigates how measurement models and statistical procedures can be applied to estimate the accuracy of proficiency classification in language testing. The paper starts with a concise introduction four models: classical test theory (CTT) model, dichotomous item response (IRT) testlet (TRT) polytomous (Poly-IRT) model. Following this, two are presented: Livingston Lewis method for CTT Rudner three IRT-based models. utility these then evaluated by examining classifying 5000...

As huge commercial value of the recommender system, there has been growing interest to improve its performance in recent years. The majority existing methods have achieved great improvement on metric click, but perform poorly conversion possibly due extremely sparse feedback signal. To track this challenge, we design a novel deep hierarchical reinforcement learning based recommendation framework model consumers' purchase interest. Specifically, high-level agent catches long-term interest,...

Energy saving and emission reduction have become common concerns in countries around the world. In China, with implementation of new strategy “carbon peak neutrality” rapid development smart grid infrastructure, amount data actual power dispatching fault analysis show exponential growth, which has led to phenomena such as poor supervision effectiveness difficulty handling faults process operation maintenance. Existing research on retrieval recommendation methods had a lower accuracy rate at...

Deep neural networks have shown promise in collaborative filtering (CF). However, existing approaches are either user-based or item-based, which cannot leverage all the underlying information explicitly. We propose CF-UIcA, a co-autoregressive model for CF tasks, exploits structural correlation domains of both users and items. The co-autoregression allows extra desired properties to be incorporated different tasks. Furthermore, we develop an efficient stochastic learning algorithm handle...

Making accurate patient care decision, as early possible, is a constant challenge, especially for physicians in the emergency department. The increasing volumes of electronic medical records (EMRs) open new horizons automatic diagnosis. In this paper, we propose to use machine learning approaches infection detection based on EMRs. Five categories information are utilized prediction, including personal information, admission note, vital signs, diagnose test results and image diagnose....

The task of next basket recommendation is pivotal for recommender systems. It involves predicting user actions, such as the product purchase or movie selection, by exploring sequential behavior and integrating users' general preferences. These elements may converge influence subsequent choices. challenge intensifies with presence varied sequences in training set, indiscriminate incorporation these can introduce superfluous noise. In response to challenges, we propose an innovative approach:...

Social networking has become a hot topic, in which recommendation algorithms are the most important. Recently, combination of deep learning and attracted considerable attention. The integration autoencoders graph convolutional neural networks, while providing an effective solution to shortcomings traditional algorithms, fails take into account user preferences risks over-smoothing as number encoder layers increases. Therefore, we introduce L1 L2 regularization techniques fuse them linearly...