Supergraph Search in Graph Databases via Hierarchical Feature-Tree

Lyu, Bingqing; Lu, Qin; Lin, Xuemin; Chang, Lijun; Yu, Jeffrey Xu

doi:10.1109/tkde.2018.2833124

Cited by 6 publications

(4 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hierarchical classification is a special case of multi-label or multi-output problems [41], [42]. It has been applied to traditional supervised learning tasks [43], such as text classification [44], [45], community data [46], case-based reasoning [47], popularity prediction [48], supergraph search in graph databases [49], road networks [50], image annotation and robot navigation. However, to the best of our knowledge, our paper is the first work successfully leveraging class hierarchy information in few-shot learning and metalearning tasks.…”

Section: Hierarchical Classificationmentioning

confidence: 99%

Many-Class Few-Shot Learning on Multi-Granularity Class Hierarchy

Liu

Zhou

Long

et al. 2022

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

We study many-class few-shot (MCFS) problem in both supervised learning and meta-learning settings. Compared to the well-studied many-class many-shot and few-class few-shot problems, the MCFS problem commonly occurs in practical applications but has been rarely studied in previous literature. It brings new challenges of distinguishing between many classes given only a few training samples per class. In this paper, we leverage the class hierarchy as a prior knowledge to train a coarse-to-fine classifier that can produce accurate predictions for MCFS problem in both settings. The propose model, "memory-augmented hierarchical-classification network (MahiNet)", performs coarse-to-fine classification where each coarse class can cover multiple fine classes. Since it is challenging to directly distinguish a variety of fine classes given few-shot data per class, MahiNet starts from learning a classifier over coarse-classes with more training data whose labels are much cheaper to obtain. The coarse classifier reduces the searching range over the fine classes and thus alleviates the challenges from "many classes". On architecture, MahiNet firstly deploys a convolutional neural network (CNN) to extract features. It then integrates a memory-augmented attention module and a multi-layer perceptron (MLP) together to produce the probabilities over coarse and fine classes. While the MLP extends the linear classifier, the attention module extends the KNN classifier, both together targeting the "few-shot" problem. We design several training strategies of MahiNet for supervised learning and meta-learning. In addition, we propose two novel benchmark datasets "mcfsImageNet" (as a subset of ImageNet) and "mcfsOmniglot" (re-splitted Omniglot) specially designed for MCFS problem. In experiments, we show that MahiNet outperforms several state-of-the-art models (e.g., prototypical networks and relation networks) on MCFS problems in both supervised learning and meta-learning.

show abstract

Section: Hierarchical Classificationmentioning

confidence: 99%

Many-Class Few-Shot Learning on Multi-Granularity Class Hierarchy

Liu

Zhou

Long

et al. 2022

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

show abstract

“…Lyu et al [24] proposes novel indexing and query processing methods and in indexing, optimum set of features is selected from the graphs. Feature tree is constructed with all sized features.…”

Section: A Work On Conventional Subgraph Miningmentioning

confidence: 99%

“…However, update of graph indexing by each successful query will be suited to improve all users' satisfaction rate. Table II • Increase in time consumption due to the execution of dynamic bag classification KNN with LSH [23] • Not suitable for large-scale graph database • High storage space is required Lightweight graph compression [24] • Results are tampered and irrelevant to the query • Index is frequently changed Lazy learning algorithm [25] • Malicious behaviors are invoked • Privacy breaches in sensitive information Top-k subgraphs [26] • Execution of top-k in each query increases Hierachical multiple clustering [27] • Not able to support large-scale graph database • Not suited for complex database KNN+MapRed uce [28] • Increases complexity while finding all nearest neighbors • Very complex method that induces the poor accuracy • Both time and space complexity is very high Peer to Peer based Indexing [32] • Time and space complexity is very high due to tree traversal • High complexity due to multiway multidimensional indexing…”

Section: Work On Query Logs Compositionmentioning

confidence: 99%

“…It is computed by the sum of time that is taken for each process during subgraph retrieval and the retrieval time 𝑄𝑅(𝑡) is computed by, 𝑄𝑅(𝑡) = 𝐴 𝑠(𝑡) + 𝐹 𝑒(𝑡) + 𝑆𝐺 𝑚(𝑡) + 𝐼 𝑚(𝑡) (24) Where 𝐴 𝑠(𝑡) is the authentication verification time, 𝐹 𝑒(𝑡) is the feature extraction time, 𝑆𝐺 𝑚(𝑡) is the subgraph mining time and 𝐼 𝑚(𝑡) is the isomorphic testing time.…”

Section: ) Query Response Timementioning

confidence: 99%

See 1 more Smart Citation

BTLA-LSDG: Blockchain-Based Triune Layered Architecture for Authenticated Subgraph Query Search in Large-Scale Dynamic Graphs

Sharmila¹,

Devi²

2023

IETE Journal of Research

View full text Add to dashboard Cite

Over the past era, subgraph mining from a large collection of graph database is a crucial problem. Existing works on subgraph mining is based on the threshold value, which returns similar graphs for a query graph. However, the number of graphs in the answer set for the same threshold can vary in terms of structure and context. In addition, scalability is another big problem due to insufficient storage.Further, it can suffer from security issues. In a distributed environment, since more queries come from different users, it is possible for the attackers to access the graph. These three problems highly exist in the current subgraph mining. To address this downside, our proposed work introduces aBlockchain-based Triune Layered Architecture for authenticated query search in the large scale dynamic graphs. The two fold process is handled in BTLA-LSDG: graph indexing and authenticated query search (query processing), which are implemented in triune layers (Data Generation Layer, Data Storage Layer, and Service Layer). Initially, data owners are authenticated to blockchain using Four-Q-Curve algorithm. The graph index is constructed by data owners and merged graph index is constructed by service providers. Based on the uploaded graph index, hash index is constructed using SHA-3. The hash index is fed into blockchain and form Dendrimer -Fractal Index. On the other hand, data user submits query with authentication. For every authenticated query, the four fold process is handled. Firstly, Multi-Constraint based Belief Entropy function is used for feature sets computation for a given query. Then Dual Similarity based MapReduce helps inmapping and reducing the relevant subgraphs with the use of optimal feature sets. Thirdly, Recurrent Neural Network (RNN)is used for subgraph isomorphic testing. Finally, graph index refinement process is undertaken to improve the query results. For this, query error is computed for each user's query at the end of retrieval. With respect to query error, fuzzy logic is used to refine the index of graph dynamically. This experiment is implemented using Hadoop environment and the results show better efficiency in terms of Scalability, Security and Storage. Further, it is tested with Precision, Recall, F-measure, Accuracy, Error Rate, Query Response Time and Positive Results.

show abstract

Optimization query algorithm of distance education database based on student behavior analysis

Gao

2023

2023 International Conference on Electronics and Devices, Computational Science (ICEDCS)

View full text Add to dashboard Cite

Supergraph Search in Graph Databases via Hierarchical Feature-Tree

Cited by 6 publications

References 27 publications

Many-Class Few-Shot Learning on Multi-Granularity Class Hierarchy

Many-Class Few-Shot Learning on Multi-Granularity Class Hierarchy

BTLA-LSDG: Blockchain-Based Triune Layered Architecture for Authenticated Subgraph Query Search in Large-Scale Dynamic Graphs

Optimization query algorithm of distance education database based on student behavior analysis

Contact Info

Product

Resources

About