GOOD-D: On Unsupervised Graph Out-Of-Distribution Detection

Liu, Yixin; Ding, Kaize; Liu, Huan; Pan, Shirui

doi:10.1145/3539597.3570446

Cited by 27 publications

(26 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Graphs are an important data structure [32][33][34][35], and GNNs are a type of method for solving graph problems that has attracted much attention recently [36,37]. Their applications span a wide range of fields, including federated learning [38], information security [39][40][41], anomaly detection [42][43][44], and the financial sector [45]. Owing to the capability of GNNs to model complex relationship nodes [46], GNN session-based recommender systems have recently become popular [47].…”

Section: Graph Neural Network For Session-based Recommendationmentioning

confidence: 99%

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

Wang,

Jin

2024

Electronics

View full text Add to dashboard Cite

Session-based recommendation plays an important role in daily life and exists in many scenarios, such as online shopping websites and streaming media platforms. Recently, some works have focused on using graph neural networks (GNNs) to recommend new items in session-based scenarios. However, these methods have encountered several limitations. First, existing methods typically ignore the impact of items’ visited time in constructing session graphs, resulting in a departure from real-world recommendation dynamics. Second, sessions are often sparse, making it challenging for GNNs to learn valuable item embedding and user preferences. Third, the existing methods usually overemphasize the impact of the last item on user preferences, neglecting their interest in multiple items in a session. To address these issues, we introduce a time-sensitive graph neural network for new item recommendation in session-based scenarios, namely, TSGNN. Specifically, TSGNN provides a novel time-sensitive session graph constructing technique to solve the first problem. For the second problem, TSGNN introduces graph augmentation and contrastive learning into it. To solve the third problem, TSGNN designs a time-aware attention mechanism to accurately discern user preferences. By evaluating the compatibility between user preferences and candidate new item embeddings, our method recommends items with high relevance scores for users. Comparative experiments demonstrate the superiority of TSGNN over state-of-the-art (SOTA) methods.

show abstract

Section: Graph Neural Network For Session-based Recommendationmentioning

confidence: 99%

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

Wang,

Jin

2024

Electronics

View full text Add to dashboard Cite

show abstract

“…Global Structure Adaptive Reconstruction. Prior research highlights the significance of using structural knowledge in graph learning (Liu et al 2023b(Liu et al , 2022c. We developed an adaptive graph topology augmentation module, G, to integrate global structural data into graph view construction.…”

Section: Global Knowledge Injected Contrastmentioning

confidence: 99%

Federated Graph Learning under Domain Shift with Generalizable Prototypes

Wan,

Huang,

2024

AAAI

View full text Add to dashboard Cite

Federated Graph Learning is a privacy-preserving collaborative approach for training a shared model on graph-structured data in the distributed environment. However, in real-world scenarios, the client graph data usually originate from diverse domains, this unavoidably hinders the generalization performance of the final global model. To address this challenge, we start the first attempt to investigate this scenario by learning a well-generalizable model. In order to improve the performance of the global model from different perspectives, we propose a novel framework called Federated Graph Learning with Generalizable Prototypes (FGGP). It decouples the global model into two levels and bridges them via prototypes. These prototypes, which are semantic centers derived from the feature extractor, can provide valuable classification information. At the classification model level, we innovatively eschew the traditional classifiers, then instead leverage clustered prototypes to capture fruitful domain information and enhance the discriminative capability of the classes, improving the performance of multi-domain predictions. Furthermore, at the feature extractor level, we go beyond traditional approaches by implicitly injecting distinct global knowledge and employing contrastive learning to obtain more powerful prototypes while enhancing the feature extractor generalization ability. Experimental results on various datasets are presented to validate the effectiveness of the proposed method.

show abstract

“…This paper examines the effect of drift on the gas dataset by detecting gas with the OOD method. ViM [38], Mahalanobis [39], KLMatching [40], MaxSoftmax [41], EnergyBased [42], MaxLogit [40], ODIN [43], OpenMax [44] and OOD methods were implemented for the experiment. AUROC was estimated from data of all positions in the 0-20 second gas leak-free section to analyze the drift effect using the OOD method.…”

Section: Drift Effect Analysis Through Out Of Distributionmentioning

confidence: 99%

A Study of Drift Effect in a Popular Metal Oxide Sensor and Gas Recognition Using Public Gas Datasets

et al. 2023

View full text Add to dashboard Cite

Metal oxide sensor is widely used in many research fields, including E-nose for gas detection due to their tunable sensitivity, space efficiency and low cost. One of the most popular open data sets in electronic nose research contains data on various gases sampled using a MOx sensor in a wind tunnel over 16 months. A recent study published in 2022 by Nik Dennler has reported the discovery of the drift effect of a public dataset due to incorrect experimental design. they reported that the order of gas collection was not randomized and further discovered that a select set of gases were collected over a particular period. This paper expands the previous paper, by analyzing the drift effect with low signal, zero-offset subtracted signal's mean, and standard deviation value by location and time, and examining it with TSNE, a dimensional reduction method. In addition, the accuracy by time and location was analyzed by applying it to various Deep Learning methods. According to the results, we confirmed that gas information is already classified before the gas leaks in terms of temporal and spatial domain. Therefore, the classification accuracy overestimates the actual accuracy that can be obtained due to the drift effect. Based on the results of this study, it is necessary to thoroughly verify the temporal and spatial validity of the gas dataset when using the publicly available gas dataset to develop gas detection algorithms.

show abstract

GOOD-D: On Unsupervised Graph Out-Of-Distribution Detection

Cited by 27 publications

References 36 publications

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

A Time-Sensitive Graph Neural Network for Session-Based New Item Recommendation

Federated Graph Learning under Domain Shift with Generalizable Prototypes

A Study of Drift Effect in a Popular Metal Oxide Sensor and Gas Recognition Using Public Gas Datasets

Contact Info

Product

Resources

About