A5087082349- Topic Modeling
- Advanced Graph Neural Networks
- Software Engineering Research
- Transportation and Mobility Innovations
- Recommender Systems and Techniques
- Risk and Safety Analysis
- Safety Systems Engineering in Autonomy
- Sharing Economy and Platforms
- Data Quality and Management
- Smart Parking Systems Research
- Software Reliability and Analysis Research
- Higher Education and Teaching Methods
- Advanced Malware Detection Techniques
- Model Reduction and Neural Networks
- Adversarial Robustness in Machine Learning
- Natural Language Processing Techniques
- Business Process Modeling and Analysis
- Image and Video Quality Assessment
- Advanced Software Engineering Methodologies
- Innovative Educational Techniques
- Imbalanced Data Classification Techniques
- Product Development and Customization
- Hydrological Forecasting Using AI
- Petri Nets in System Modeling
- Nuclear Physics and Applications
Beihang University
2014-2024
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.
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...
The extraction of Metal-Organic Frameworks (MOFs) synthesis conditions from literature text has been challenging but crucial for the logical design new MOFs with desirable functionality. recent advent large language models (LLMs) provides disruptively solution to this long-standing problem and latest researches have reported over 90% F1 in extracting correct literature. We argue paper that most existing practices LLMs stay primitive zero-shot learning, which could lead downgraded application...
Software defect (Bug) prediction plays an important role in improving software quality. Many approaches have been proposed and achieved great effects the real-world. However, existing works are usually constrained only one project, hence their effectiveness on cross-project (cross-prediction) is poor. This mainly because of problem class imbalance feature distribution differences between source target projects. In this paper, we effective method called Transfer Component Analysis Neural...
Neural translation model has greatly grown in recent years. Many researches have come up with very good solutions to deficiencies model. However, it is difficult get best effect for rare words and terminologies what are marked as unknown because of the limit dictionary's size. This paper presents a bidirectional can be used translate between bilinguals optimize terminologies. At first we use word2vec word similarity By replacing trained tested by model, solve problems caused words. In...
An important characteristic of a fractal signal is that its graph not smooth in any small interval. This indicates the difficulty approximation signals, because traditional methods normally require some certain smoothness approximated function. However, recent studies have shown functions satisfy Hölder condition can be linearly changed dimension their graphs by fractional calculus, which implies we use calculus to make signals smoother, and then approximate these signals. paper first gives...
In this paper, we present a model named VuLASTE, treating vulnerability detection as specialized text classification task. To address the vocabulary explosion problem, VuLASTE utilizes byte-level BPE algorithm from natural language processing. Within introduce novel AST path embedding to represent source code nesting information. Additionally, employ combination of global and dilated window attention Longformer extract long sequence semantics code. tackle issue data imbalance, common...
In real-time and embedded systems, time is a very important feature, time-triggered architecture strong candidate platform. However, because UML built on event-triggered mechanisms, it not suitable for mechanisms. This paper introduces profile MARTE, presents some modeling mechanisms which describe the features based MARTE. Finally, also elaborates application of these through real case.
The name disambiguation task is designed to solve the ambiguity problem of documents multiple persons who have same with one another. aims partition all publications belonging person and realize that each decomposed compos ed a unique person. Many works on common feature clustering method usually used in last step. paper presents complementary study these from another point view. Based idea strong association relationships are likely belong author, this proposes discovering meta clusters by...
To address the data sparsity and cold start issues of collaborative filtering, side information, such as social network, knowledge graph, is introduced to recommender systems. Knowledge a sort auxiliary structural data, full semantic logical connections among entities in world. In this paper, we propose Hierarchical Interest Propagation Network(HKIPN) for recommendation, where new heterogeneous propagation method presented. Specifically, HKIPN propagates user interest simultaneously unified...
After analyzing the shortcomings of current researches combining Petri Net with safety analysis, this paper proposed a system analysis method based on stochastic Time Nets. System model built by Nets is neither limited to exponential and deterministic transitions nor enabling restriction for generally distributed transitions. Steady-State Safety path measurements can be calculated through generating state classes graph, transient Markov theory. Finally, an application example given show...
Mining the evolutionary rules of source code files can be conducted by analyzing data generated in development open software. In this paper, log information two famous projects is collected and statistical distribution number developers corresponding to class modification analyzed method. As a result, we discover that fellows approximately an exponential distribution. addition, analyze features function structure kinds both developed who have too many behaviors their modified tend more...
This study conducts a data-driven statistic analysis for comparatively examining early programmers' academic performance and learning behavior between SPOC blended teaching class traditional in an introductory C programming course. Teaching activities the experimental are implemented by supporting tool -"rainy class", while control group implements classroom teaching. Both two groups required to complete assigned labs automatic assessment system-the online judge, log data terms of students'...
Many researchers have studied optimization methods for ridesharing. However, the individual interests of passengers and drivers are not considered enough. So we propose a two-sided stable matching method according to actual preferences passengers(requesters) drivers(workers). We also design pruning algorithm based on Euclidean distance speed up process. Experiments real data show that our can perform well.
Automatic Term Extraction is a common upstream task of both natural language and programming processing tasks. However, most existing domain-specific term extraction methods are designed for language, do not satisfy demand language. In this paper we proposed an automated cross-language method extracting terms from source code. To test the on real software projects, choose open repositories in Github Collections. Then results our TF-IDF were checked by crowdsourcing, with 12 experienced...
The water pollution level is affected by many factors that may have different degree in rivers.The factor weighting model combining the analysis with fuzzy mathematical can advantages of both sides.The results based on simulated data set shows this new algorithm provides an effective evaluation.And thought to provide a common method for quality evaluation.