Inhaltsverzeichnis
Auswahl der wissenschaftlichen Literatur zum Thema „Graph-based input representation“
Geben Sie eine Quelle nach APA, MLA, Chicago, Harvard und anderen Zitierweisen an
Machen Sie sich mit den Listen der aktuellen Artikel, Bücher, Dissertationen, Berichten und anderer wissenschaftlichen Quellen zum Thema "Graph-based input representation" bekannt.
Neben jedem Werk im Literaturverzeichnis ist die Option "Zur Bibliographie hinzufügen" verfügbar. Nutzen Sie sie, wird Ihre bibliographische Angabe des gewählten Werkes nach der nötigen Zitierweise (APA, MLA, Harvard, Chicago, Vancouver usw.) automatisch gestaltet.
Sie können auch den vollen Text der wissenschaftlichen Publikation im PDF-Format herunterladen und eine Online-Annotation der Arbeit lesen, wenn die relevanten Parameter in den Metadaten verfügbar sind.
Zeitschriftenartikel zum Thema "Graph-based input representation"
1
Lu, Fangbo, Zhihao Zhang und Changsheng Shui. „Online trajectory anomaly detection model based on graph neural networks and variational autoencoder“. Journal of Physics: Conference Series 2816, Nr. 1 (01.08.2024): 012006. http://dx.doi.org/10.1088/1742-6596/2816/1/012006.
Der volle Inhalt der QuelleAnnotation:
Abstract To efficiently determine whether an entire trajectory exhibits abnormal behavior, we introduce an online trajectory anomaly detection model known as GeoGNFTOD, which employs graph neural networks for road segment representation, creating a directed graph by mapping trajectories onto the road network. The graph representation is constructed based on the road segments in this directed graph. By utilizing Transformer sequence encoding, the trajectory representation is derived and hierarchical geographic encoding captures the GPS mapping of the original trajectories. Merging these two representations produces the final trajectory representation, serving as input for an LSTM-based variational autoencoder to reconstruct the trajectory representation, facilitating rapid online anomaly detection. Experimental findings on a large-scale taxi trajectory dataset illustrate the superior performance of GeoGNFTOD compared to baseline algorithms.
2
Yu, Xingtong, Zemin Liu, Yuan Fang und Xinming Zhang. „Learning to Count Isomorphisms with Graph Neural Networks“. Proceedings of the AAAI Conference on Artificial Intelligence 37, Nr. 4 (26.06.2023): 4845–53. http://dx.doi.org/10.1609/aaai.v37i4.25610.
Der volle Inhalt der QuelleAnnotation:
Subgraph isomorphism counting is an important problem on graphs, as many graph-based tasks exploit recurring subgraph patterns. Classical methods usually boil down to a backtracking framework that needs to navigate a huge search space with prohibitive computational cost. Some recent studies resort to graph neural networks (GNNs) to learn a low-dimensional representation for both the query and input graphs, in order to predict the number of subgraph isomorphisms on the input graph. However, typical GNNs employ a node-centric message passing scheme that receives and aggregates messages on nodes, which is inadequate in complex structure matching for isomorphism counting. Moreover, on an input graph, the space of possible query graphs is enormous, and different parts of the input graph will be triggered to match different queries. Thus, expecting a fixed representation of the input graph to match diversely structured query graphs is unrealistic. In this paper, we propose a novel GNN called Count-GNN for subgraph isomorphism counting, to deal with the above challenges. At the edge level, given that an edge is an atomic unit of encoding graph structures, we propose an edge-centric message passing scheme, where messages on edges are propagated and aggregated based on the edge adjacency to preserve fine-grained structural information. At the graph level, we modulate the input graph representation conditioned on the query, so that the input graph can be adapted to each query individually to improve their matching. Finally, we conduct extensive experiments on a number of benchmark datasets to demonstrate the superior performance of Count-GNN.
3
Bauer, Daniel. „Understanding Descriptions of Visual Scenes Using Graph Grammars“. Proceedings of the AAAI Conference on Artificial Intelligence 27, Nr. 1 (29.06.2013): 1656–57. http://dx.doi.org/10.1609/aaai.v27i1.8498.
Der volle Inhalt der QuelleAnnotation:
Automatic generation of 3D scenes from descriptions has applications in communication, education, and entertainment, but requires deep understanding of the input text. I propose thesis work on language understanding using graph-based meaning representations that can be decomposed into primitive spatial relations. The techniques used for analyzing text and transforming it into a scene representation are based on context-free graph grammars. The thesis develops methods for semantic parsing with graphs, acquisition of graph grammars, and satisfaction of spatial and world-knowledge constraints during parsing.
4
Wu, Xinyue, und Huilin Chen. „Augmented Feature Diffusion on Sparsely Sampled Subgraph“. Electronics 13, Nr. 16 (15.08.2024): 3249. http://dx.doi.org/10.3390/electronics13163249.
Der volle Inhalt der QuelleAnnotation:
Link prediction is a fundamental problem in graphs. Currently, SubGraph Representation Learning (SGRL) methods provide state-of-the-art solutions for link prediction by transforming the task into a graph classification problem. However, existing SGRL solutions suffer from high computational costs and lack scalability. In this paper, we propose a novel SGRL framework called Augmented Feature Diffusion on Sparsely Sampled Subgraph (AFD3S). The AFD3S first uses a conditional variational autoencoder to augment the local features of the input graph, effectively improving the expressive ability of downstream Graph Neural Networks. Then, based on a random walk strategy, sparsely sampled subgraphs are obtained from the target node pairs, reducing computational and storage overhead. Graph diffusion is then performed on the sampled subgraph to achieve specific weighting. Finally, the diffusion matrix of the subgraph and its augmented feature matrix are used for feature diffusion to obtain operator-level node representations as inputs for the SGRL-based link prediction. Feature diffusion effectively simulates the message-passing process, simplifying subgraph representation learning, thus accelerating the training and inference speed of subgraph learning. Our proposed AFD3S achieves optimal prediction performance on several benchmark datasets, with significantly reduced storage and computational costs.
5
Cooray, Thilini, und Ngai-Man Cheung. „Graph-Wise Common Latent Factor Extraction for Unsupervised Graph Representation Learning“. Proceedings of the AAAI Conference on Artificial Intelligence 36, Nr. 6 (28.06.2022): 6420–28. http://dx.doi.org/10.1609/aaai.v36i6.20593.
Der volle Inhalt der QuelleAnnotation:
Unsupervised graph-level representation learning plays a crucial role in a variety of tasks such as molecular property prediction and community analysis, especially when data annotation is expensive. Currently, most of the best-performing graph embedding methods are based on Infomax principle. The performance of these methods highly depends on the selection of negative samples and hurt the performance, if the samples were not carefully selected. Inter-graph similarity-based methods also suffer if the selected set of graphs for similarity matching is low in quality. To address this, we focus only on utilizing the current input graph for embedding learning. We are motivated by an observation from real-world graph generation processes where the graphs are formed based on one or more global factors which are common to all elements of the graph (e.g., topic of a discussion thread, solubility level of a molecule). We hypothesize extracting these common factors could be highly beneficial. Hence, this work proposes a new principle for unsupervised graph representation learning: Graph-wise Common latent Factor EXtraction (GCFX). We further propose a deep model for GCFX, deepGCFX, based on the idea of reversing the above-mentioned graph generation process which could explicitly extract common latent factors from an input graph and achieve improved results on downstream tasks to the current state-of-the-art. Through extensive experiments and analysis, we demonstrate that, while extracting common latent factors is beneficial for graph-level tasks to alleviate distractions caused by local variations of individual nodes or local neighbourhoods, it also benefits node-level tasks by enabling long-range node dependencies, especially for disassortative graphs.
6
Gildea, Daniel, Giorgio Satta und Xiaochang Peng. „Ordered Tree Decomposition for HRG Rule Extraction“. Computational Linguistics 45, Nr. 2 (Juni 2019): 339–79. http://dx.doi.org/10.1162/coli_a_00350.
Der volle Inhalt der QuelleAnnotation:
We present algorithms for extracting Hyperedge Replacement Grammar (HRG) rules from a graph along with a vertex order. Our algorithms are based on finding a tree decomposition of smallest width, relative to the vertex order, and then extracting one rule for each node in this structure. The assumption of a fixed order for the vertices of the input graph makes it possible to solve the problem in polynomial time, in contrast to the fact that the problem of finding optimal tree decompositions for a graph is NP-hard. We also present polynomial-time algorithms for parsing based on our HRGs, where the input is a vertex sequence and the output is a graph structure. The intended application of our algorithms is grammar extraction and parsing for semantic representation of natural language. We apply our algorithms to data annotated with Abstract Meaning Representations and report on the characteristics of the resulting grammars.
7
Miao, Fengyu, Xiuzhuang Zhou, Shungen Xiao und Shiliang Zhang. „A Graph Similarity Algorithm Based on Graph Partitioning and Attention Mechanism“. Electronics 13, Nr. 19 (25.09.2024): 3794. http://dx.doi.org/10.3390/electronics13193794.
Der volle Inhalt der QuelleAnnotation:
In recent years, graph similarity algorithms have been extensively developed based on neural networks. However, with an increase in the node count in graphs, these models either suffer from a reduced representation ability or face a significant increase in the computational cost. To address this issue, a graph similarity algorithm based on graph partitioning and attention mechanisms was proposed in this study. Our method first divided each input graph into the subgraphs to directly extract the local structural features. The residual graph convolution and multihead self-attention mechanisms were employed to generate node embeddings for each subgraph, extract the feature information from the nodes, and regenerate the subgraph embeddings using varying attention weights. Initially, rough cosine similarity calculations were performed on all subgraph pairs from the two sets of subgraphs, with highly similar pairs selected for precise similarity computation. These results were then integrated into the similarity score for the input graph. The experimental results indicated that the proposed learning algorithm outperformed the traditional algorithms and similar computing models in terms of graph similarity computation performance.
8
Coşkun, Kemal Çağlar, Muhammad Hassan und Rolf Drechsler. „Equivalence Checking of System-Level and SPICE-Level Models of Linear Circuits“. Chips 1, Nr. 1 (13.06.2022): 54–71. http://dx.doi.org/10.3390/chips1010006.
Der volle Inhalt der QuelleAnnotation:
Due to the increasing complexity of analog circuits and their integration into System-on-Chips (SoC), the analog design and verification industry would greatly benefit from an expansion of system-level methodologies using SystemC AMS. These can provide a speed increase of over 100,000× in comparison to SPICE-level simulations and allow interoperability with digital tools at the system-level. However, a key barrier to the expansion of system-level tools for analog circuits is the lack of confidence in system-level models implemented in SystemC AMS. Functional equivalence of single Laplace Transfer Function (LTF) system-level models to respective SPICE-level models was successfully demonstrated recently. However, this is clearly not sufficient, as the complex systems comprise multiple LTF modules. In this article, we go beyond single LTF models, i.e., we develop a novel graph-based methodology to formally check equivalence between complex system-level and SPICE-level representations of Single-Input Single-Output (SISO) linear analog circuits, such as High-Pass Filters (HPF). To achieve this, first, we introduce a canonical representation in the form of a Signal-Flow Graph (SFG), which is used to functionally map the two representations from separate modeling levels. This canonical representation consists of the input and output nodes and a single edge between them with an LTF as its weight. Second, we create an SFG representation with linear graph modeling for SPICE-level models, whereas for system-level models we extract an SFG from the behavioral description. We then transform the SFG representations into the canonical representation by utilizing three graph manipulation techniques, namely node removal, parallel edge unification, and reflexive edge elimination. This allows us to establish functional equivalence between complex system-level models and SPICE-level models. We demonstrate the applicability of the proposed methodology by successfully applying it to complex circuits.
9
Zhang, Dong, Suzhong Wei, Shoushan Li, Hanqian Wu, Qiaoming Zhu und Guodong Zhou. „Multi-modal Graph Fusion for Named Entity Recognition with Targeted Visual Guidance“. Proceedings of the AAAI Conference on Artificial Intelligence 35, Nr. 16 (18.05.2021): 14347–55. http://dx.doi.org/10.1609/aaai.v35i16.17687.
Der volle Inhalt der QuelleAnnotation:
Multi-modal named entity recognition (MNER) aims to discover named entities in free text and classify them into pre-defined types with images. However, dominant MNER models do not fully exploit fine-grained semantic correspondences between semantic units of different modalities, which have the potential to refine multi-modal representation learning. To deal with this issue, we propose a unified multi-modal graph fusion (UMGF) approach for MNER. Specifically, we first represent the input sentence and image using a unified multi-modal graph, which captures various semantic relationships between multi-modal semantic units (words and visual objects). Then, we stack multiple graph-based multi-modal fusion layers that iteratively perform semantic interactions to learn node representations. Finally, we achieve an attention-based multi-modal representation for each word and perform entity labeling with a CRF decoder. Experimentation on the two benchmark datasets demonstrates the superiority of our MNER model.
10
Ren, Min, Yunlong Wang, Zhenan Sun und Tieniu Tan. „Dynamic Graph Representation for Occlusion Handling in Biometrics“. Proceedings of the AAAI Conference on Artificial Intelligence 34, Nr. 07 (03.04.2020): 11940–47. http://dx.doi.org/10.1609/aaai.v34i07.6869.
Der volle Inhalt der QuelleAnnotation:
The generalization ability of Convolutional neural networks (CNNs) for biometrics drops greatly due to the adverse effects of various occlusions. To this end, we propose a novel unified framework integrated the merits of both CNNs and graphical models to learn dynamic graph representations for occlusion problems in biometrics, called Dynamic Graph Representation (DGR). Convolutional features onto certain regions are re-crafted by a graph generator to establish the connections among the spatial parts of biometrics and build Feature Graphs based on these node representations. Each node of Feature Graphs corresponds to a specific part of the input image and the edges express the spatial relationships between parts. By analyzing the similarities between the nodes, the framework is able to adaptively remove the nodes representing the occluded parts. During dynamic graph matching, we propose a novel strategy to measure the distances of both nodes and adjacent matrixes. In this way, the proposed method is more convincing than CNNs-based methods because the dynamic graph method implies a more illustrative and reasonable inference of the biometrics decision. Experiments conducted on iris and face demonstrate the superiority of the proposed framework, which boosts the accuracy of occluded biometrics recognition by a large margin comparing with baseline methods.
Dissertationen zum Thema "Graph-based input representation"
1
Agarwal, Navneet. „Autοmated depressiοn level estimatiοn : a study οn discοurse structure, input representatiοn and clinical reliability“. Electronic Thesis or Diss., Normandie, 2024. http://www.theses.fr/2024NORMC215.
Der volle Inhalt der QuelleAnnotation:
Compte tenu de l'impact sévère et généralisé de la dépression, des initiatives de recherche significatives ont été entreprises pour définir des systèmes d'évaluation automatisée de la dépression. La recherche présentée dans cette thèse tourne autour des questions suivantes qui restent relativement inexplorées malgré leur pertinence dans le domaine de l'évaluation automatisée de la dépression : (1) le rôle de la structure du discours dans l'analyse de la santé mentale, (2) la pertinence de la représentation de l'entrée pour les capacités prédictives des modèles de réseaux neuronaux, et (3) l'importance de l'expertise du domaine dans la détection automatisée de la dépression.La nature dyadique des entretiens patient-thérapeute garantit la présence d'une structure sous-jacente complexe dans le discours. Dans cette thèse, nous établissons d'abord l'importance des questions du thérapeute dans l'entrée du modèle de réseau neuronal, avant de montrer qu'une combinaison séquentielle des entrées du patient et du thérapeute est une stratégie sous-optimale. Par conséquent, des architectures à vues multiples sont proposées comme moyen d'incorporer la structure du discours dans le processus d'apprentissage des réseaux neuronaux. Les résultats expérimentaux obtenus avec deux encodages de texte différents montrent les avantages des architectures multi-vues proposées, validant la pertinence de la conservation de la structure du discours dans le processus d'apprentissage du modèle.Ayant établi la nécessité de conserver la structure du discours dans le processus d'apprentissage, nous explorons plus avant les représentations textuelles basées sur les graphes. Les recherches menées dans ce contexte mettent en évidence l'impact des représentations d'entrée non seulement pour définir les capacités d'apprentissage du modèle, mais aussi pour comprendre leur processus prédictif. Les graphiques de similitude de phrases et les graphiques de corrélation de mots-clés sont utilisés pour illustrer la capacité des représentations graphiques à fournir des perspectives variées sur la même entrée, en mettant en évidence des informations qui peuvent non seulement améliorer les performances prédictives des modèles, mais aussi être pertinentes pour les professionnels de la santé. Le concept de vues multiples est également incorporé dans les deux structures graphiques afin de mettre en évidence les différences de perspectives entre le patient et le thérapeute au cours d'un même entretien. En outre, il est démontré que la visualisation des structures graphiques proposées peut fournir des informations précieuses indiquant des changements subtils dans le comportement du patient et du thérapeute, faisant allusion à l'état mental du patient.Enfin, nous soulignons le manque d'implication des professionnels de la santé dans le contexte de la détection automatique de la dépression basée sur des entretiens cliniques. Dans le cadre de cette thèse, des annotations cliniques de l'ensemble de données DAIC-WOZ ont été réalisées afin de fournir une ressource pour mener des recherches interdisciplinaires dans ce domaine. Des expériences sont définies pour étudier l'intégration des annotations cliniques dans les modèles de réseaux neuronaux appliqués à la tâche de prédiction au niveau des symptômes dans le domaine de la détection automatique de la dépression. En outre, les modèles proposés sont analysés dans le contexte des annotations cliniques afin d'établir une analogie entre leur processus prédictif et leurs tendances psychologiques et ceux des professionnels de la santé, ce qui constitue une étape vers l'établissement de ces modèles en tant qu'outils cliniques fiablesGiven the severe and widespread impact of depression, significant research initiatives have been undertaken to define systems for automated depression assessment. The research presented in this dissertation revolves around the following questions that remain relatively unexplored despite their relevance within automated depression assessment domain; (1) the role of discourse structure in mental health analysis, (2) the relevance of input representation towards the predictive abilities of neural network models, and (3) the importance of domain expertise in automated depression detection.The dyadic nature of patient-therapist interviews ensures the presence of a complex underlying structure within the discourse. Within this thesis, we first establish the importance of therapist questions within the neural network model's input, before showing that a sequential combination of patient and therapist input is a sub-optimal strategy. Consequently, Multi-view architectures are proposed as a means of incorporating the discourse structure within the learning process of neural networks. Experimental results with two different text encodings show the advantages of the proposed multi-view architectures, validating the relevance of retaining discourse structure within the model's training process.Having established the need to retain the discourse structure within the learning process, we further explore graph based text representations. The research conducted in this context highlights the impact of input representations not only in defining the learning abilities of the model, but also in understanding their predictive process. Sentence Similarity Graphs and Keyword Correlation Graphs are used to exemplify the ability of graphical representations to provide varying perspectives of the same input, highlighting information that can not only improve the predictive performance of the models but can also be relevant for medical professionals. Multi-view concept is also incorporated within the two graph structures to further highlight the difference in the perspectives of the patient and the therapist within the same interview. Furthermore, it is shown that visualization of the proposed graph structures can provide valuable insights indicative of subtle changes in patient and therapist's behavior, hinting towards the mental state of the patient.Finally, we highlight the lack of involvement of medical professionals within the context of automated depression detection based on clinical interviews. As part of this thesis, clinical annotations of the DAIC-WOZ dataset were performed to provide a resource for conducting interdisciplinary research in this field. Experiments are defined to study the integration of the clinical annotations within the neural network models applied to symptom-level prediction task within the automated depression detection domain. Furthermore, the proposed models are analyzed in the context of the clinical annotations to analogize their predictive process and psychological tendencies with those of medical professionals, a step towards establishing them as reliable clinical tools
Buchteile zum Thema "Graph-based input representation"
1
Jagan, Balaji, Ranjani Parthasarathi und Geetha T. V. „Graph-Based Abstractive Summarization“. In Innovations, Developments, and Applications of Semantic Web and Information Systems, 236–61. IGI Global, 2018. http://dx.doi.org/10.4018/978-1-5225-5042-6.ch009.
Der volle Inhalt der QuelleAnnotation:
Customization of information from web documents is an immense job that involves mainly the shortening of original texts. Extractive methods use surface level and statistical features for the selection of important sentences. In contrast, abstractive methods need a formal semantic representation, where the selection of important components and the rephrasing of the selected components are carried out using the semantic features associated with the words as well as the context. In this paper, we propose a semi-supervised bootstrapping approach for the identification of important components for abstractive summarization. The input to the proposed approach is a fully connected semantic graph of a document, where the semantic graphs are constructed for sentences, which are then connected by synonym concepts and co-referring entities to form a complete semantic graph. The direction of the traversal of nodes is determined by a modified spreading activation algorithm, where the importance of the nodes and edges are decided, based on the node and its connected edges under consideration.
2
Kumar, P. Krishna, und Harish G. Ramaswamy. „Graph Classification with GNNs: Optimisation, Representation & Inductive Bias“. In Frontiers in Artificial Intelligence and Applications. IOS Press, 2024. http://dx.doi.org/10.3233/faia240726.
Der volle Inhalt der QuelleAnnotation:
Theoretical studies on the representation power of GNNs have been centered around understanding the equivalence of GNNs, using WL-Tests for detecting graph isomorphism. In this paper, we argue that such equivalence ignores the accompanying optimization issues and does not provide a holistic view of the GNN learning process. We illustrate these gaps between representation and optimization with examples and experiments. We also explore the existence of an implicit inductive bias (e.g. fully connected networks prefer to learn low frequency functions in their input space) in GNNs, in the context of graph classification tasks. We further prove theoretically that the message-passing layers in the graph, have a tendency to search for either discriminative subgraphs, or a collection of discriminative nodes dispersed across the graph, depending on the different global pooling layers used. We empirically verify this bias through experiments over real-world and synthetic datasets. Finally, we show how our work can help in incorporating domain knowledge via attention based architectures, and can evince their capability to discriminate coherent subgraphs.
3
Toropov, Andrey A., Alla P. Toropova, Emilio Benfenati, Orazio Nicolotti, Angelo Carotti, Karel Nesmerak, Aleksandar M. Veselinović et al. „QSPR/QSAR Analyses by Means of the CORAL Software“. In Pharmaceutical Sciences, 929–55. IGI Global, 2017. http://dx.doi.org/10.4018/978-1-5225-1762-7.ch036.
Der volle Inhalt der QuelleAnnotation:
In this chapter, the methodology of building up quantitative structure—property/activity relationships (QSPRs/QSARs)—by means of the CORAL software is described. The Monte Carlo method is the basis of this approach. Simplified Molecular Input-Line Entry System (SMILES) is used as the representation of the molecular structure. The conversion of SMILES into the molecular graph is available for QSPR/QSAR analysis using the CORAL software. The model for an endpoint is a mathematical function of the correlation weights for various features of the molecular structure. Hybrid models that are based on features extracted from both SMILES and a graph also can be built up by the CORAL software. The conceptually new ideas collected and revealed through the CORAL software are: (1) any QSPR/QSAR model is a random event; and (2) optimal descriptor can be a translator of eclectic information into an endpoint prediction.
4
Toropov, Andrey A., Alla P. Toropova, Emilio Benfenati, Orazio Nicolotti, Angelo Carotti, Karel Nesmerak, Aleksandar M. Veselinović et al. „QSPR/QSAR Analyses by Means of the CORAL Software“. In Quantitative Structure-Activity Relationships in Drug Design, Predictive Toxicology, and Risk Assessment, 560–85. IGI Global, 2015. http://dx.doi.org/10.4018/978-1-4666-8136-1.ch015.
Der volle Inhalt der QuelleAnnotation:
In this chapter, the methodology of building up quantitative structure—property/activity relationships (QSPRs/QSARs)—by means of the CORAL software is described. The Monte Carlo method is the basis of this approach. Simplified Molecular Input-Line Entry System (SMILES) is used as the representation of the molecular structure. The conversion of SMILES into the molecular graph is available for QSPR/QSAR analysis using the CORAL software. The model for an endpoint is a mathematical function of the correlation weights for various features of the molecular structure. Hybrid models that are based on features extracted from both SMILES and a graph also can be built up by the CORAL software. The conceptually new ideas collected and revealed through the CORAL software are: (1) any QSPR/QSAR model is a random event; and (2) optimal descriptor can be a translator of eclectic information into an endpoint prediction.
5
Zhang, Taolin, Dongyang Li, Qizhou Chen, Chengyu Wang, Longtao Huang, Hui Xue, Xiaofeng He und Jun Huang. „R4: Reinforced Retriever-Reorder-Responder for Retrieval-Augmented Large Language Models“. In Frontiers in Artificial Intelligence and Applications. IOS Press, 2024. http://dx.doi.org/10.3233/faia240755.
Der volle Inhalt der QuelleAnnotation:
Retrieval-augmented large language models (LLMs) leverage relevant content retrieved by information retrieval systems to generate correct responses, aiming to alleviate the hallucination problem. However, existing retriever-responder methods typically append relevant documents to the prompt of LLMs to perform text generation tasks without considering the interaction of fine-grained structural semantics between the retrieved documents and the LLMs. This issue is particularly important for accurate response generation as LLMs tend to “lose in the middle” when dealing with input prompts augmented with lengthy documents. In this work, we propose a new pipeline named “Reinforced Retriever-Reorder-Responder” (R4) to learn document orderings for retrieval-augmented LLMs, thereby further enhancing their generation abilities while the large numbers of parameters of LLMs remain frozen. The reordering learning process is divided into two steps according to the quality of the generated responses: document order adjustment and document representation enhancement. Specifically, document order adjustment aims to organize retrieved document orderings into beginning, middle, and end positions based on graph attention learning, which maximizes the reinforced reward of response quality. Document representation enhancement further refines the representations of retrieved documents for responses of poor quality via document-level gradient adversarial learning. Extensive experiments demonstrate that our proposed pipeline achieves better factual question-answering performance on knowledge-intensive tasks compared to strong baselines across various public datasets. The source codes and trained models will be released upon paper acceptance.
6
Yang, Zixuan, Xiao Wang, Yanhua Yu, Yuling Wang, Kangkang Lu, Zirui Guo, Xiting Qin, Yunshan Ma und Tat-Seng Chua. „Hop-based Heterogeneous Graph Transformer“. In Frontiers in Artificial Intelligence and Applications. IOS Press, 2024. http://dx.doi.org/10.3233/faia240759.
Der volle Inhalt der QuelleAnnotation:
The Graph Transformer (GT) has shown significant ability in processing graph-structured data, addressing limitations in graph neural networks, such as over-smoothing and over-squashing. However, the implementation of GT in real-world heterogeneous graphs (HGs) with complex topology continues to present numerous challenges. Firstly, a challenge arises in designing a tokenizer that is compatible with heterogeneity. Secondly, the complexity of the transformer hampers the acquisition of high-order neighbor information in HGs. In this paper, we propose a novel Hop-based Heterogeneous Graph Transformer (H2Gormer) framework, paving a promising path for HGs to benefit from the capabilities of Transformers. We propose a Heterogeneous Hop-based Token Generation module to obtain high-order information in a flexible way. Specifically, to enrich the fine-grained heterogeneous semantics of each token, we propose a tailored multi-relational encoder to encode the hop-based neighbors. In this way, the resulting token embeddings are input to the Hop-based Transformer to obtain node representations, which are then combined with position embeddings to obtain the final encoding. Extensive experiments on four datasets are conducted to demonstrate the effectiveness of H2Gormer.
7
Omerovic, Aida, Amela Karahasanovic und Ketil Stølen. „Uncertainty Handling in Weighted Dependency Trees“. In Dependability and Computer Engineering, 381–416. IGI Global, 2012. http://dx.doi.org/10.4018/978-1-60960-747-0.ch016.
Der volle Inhalt der QuelleAnnotation:
Weighted dependency trees (WDTs) are used in a multitude of approaches to system analysis, such as fault tree analysis or event tree analysis. In fact, any acyclic graph can be transformed to a WDT. Important decisions are often based on WDT analysis. Common for all WDT-based approaches is the inherent uncertainty due to lack or inaccuracy of the input data. In order to indicate credibility of such WDT analysis, uncertainty handling is essential. There is however, to our knowledge, no comprehensive evaluation of the uncertainty handling approaches in the context of the WDTs. This chapter aims to rectify this. We concentrate on approaches applicable for epistemic uncertainty related to empirical input. The existing and the potentially useful approaches are identified through a systematic literature review. The approaches are then outlined and evaluated at a high-level, before a restricted set undergoes a more detailed evaluation based on a set of pre-defined evaluation criteria. We argue that the epistemic uncertainty is better suited for possibilistic uncertainty representations than the probabilistic ones. The results indicate that precision, expressiveness, predictive accuracy, scalability on real-life systems, and comprehensibility are among the properties which differentiate the approaches. The selection of a preferred approach should depend on the degree of need for certain properties relative to others, given the context. The right trade off is particularly important when the input is based on both expert judgments and measurements. The chapter may serve as a roadmap for examining the uncertainty handling approaches, or as a resource for identifying the adequate one.
Konferenzberichte zum Thema "Graph-based input representation"
1
Morris, Matthew, David J. Tena Cucala, Bernardo Cuenca Grau und Ian Horrocks. „Relational Graph Convolutional Networks Do Not Learn Sound Rules“. In 21st International Conference on Principles of Knowledge Representation and Reasoning {KR-2023}, 897–908. California: International Joint Conferences on Artificial Intelligence Organization, 2024. http://dx.doi.org/10.24963/kr.2024/84.
Der volle Inhalt der QuelleAnnotation:
Graph neural networks (GNNs) are frequently used to predict missing facts in knowledge graphs (KGs). Motivated by the lack of explainability for the outputs of these models, recent work has aimed to explain their predictions using Datalog, a widely used logic-based formalism. However, such work has been restricted to certain subclasses of GNNs. In this paper, we consider one of the most popular GNN architectures for KGs, R-GCN, and we provide two methods to extract rules that explain its predictions and are sound, in the sense that each fact derived by the rules is also predicted by the GNN, for any input dataset. Furthermore, we provide a method that can verify that certain classes of Datalog rules are not sound for the R-GCN. In our experiments, we train R-GCNs on KG completion benchmarks, and we are able to verify that no Datalog rule is sound for these models, even though the models often obtain high to near-perfect accuracy. This raises some concerns about the ability of R-GCN models to generalise and about the explainability of their predictions. We further provide two variations to the training paradigm of R-GCN that encourage it to learn sound rules and find a trade-off between model accuracy and the number of learned sound rules.
2
Guo, Zhichun, Kehan Guo, Bozhao Nan, Yijun Tian, Roshni G. Iyer, Yihong Ma, Olaf Wiest et al. „Graph-based Molecular Representation Learning“. In Thirty-Second International Joint Conference on Artificial Intelligence {IJCAI-23}. California: International Joint Conferences on Artificial Intelligence Organization, 2023. http://dx.doi.org/10.24963/ijcai.2023/744.
Der volle Inhalt der QuelleAnnotation:
Molecular representation learning (MRL) is a key step to build the connection between machine learning and chemical science. In particular, it encodes molecules as numerical vectors preserving the molecular structures and features, on top of which the downstream tasks (e.g., property prediction) can be performed. Recently, MRL has achieved considerable progress, especially in methods based on deep molecular graph learning. In this survey, we systematically review these graph-based molecular representation techniques, especially the methods incorporating chemical domain knowledge. Specifically, we first introduce the features of 2D and 3D molecular graphs. Then we summarize and categorize MRL methods into three groups based on their input. Furthermore, we discuss some typical chemical applications supported by MRL. To facilitate studies in this fast-developing area, we also list the benchmarks and commonly used datasets in the paper. Finally, we share our thoughts on future research directions.
3
Jin, Ming, Yizhen Zheng, Yuan-Fang Li, Chen Gong, Chuan Zhou und Shirui Pan. „Multi-Scale Contrastive Siamese Networks for Self-Supervised Graph Representation Learning“. In Thirtieth International Joint Conference on Artificial Intelligence {IJCAI-21}. California: International Joint Conferences on Artificial Intelligence Organization, 2021. http://dx.doi.org/10.24963/ijcai.2021/204.
Der volle Inhalt der QuelleAnnotation:
Graph representation learning plays a vital role in processing graph-structured data. However, prior arts on graph representation learning heavily rely on labeling information. To overcome this problem, inspired by the recent success of graph contrastive learning and Siamese networks in visual representation learning, we propose a novel self-supervised approach in this paper to learn node representations by enhancing Siamese self-distillation with multi-scale contrastive learning. Specifically, we first generate two augmented views from the input graph based on local and global perspectives. Then, we employ two objectives called cross-view and cross-network contrastiveness to maximize the agreement between node representations across different views and networks. To demonstrate the effectiveness of our approach, we perform empirical experiments on five real-world datasets. Our method not only achieves new state-of-the-art results but also surpasses some semi-supervised counterparts by large margins. Code is made available at https://github.com/GRAND-Lab/MERIT
4
Jin, Di, Luzhi Wang, Yizhen Zheng, Xiang Li, Fei Jiang, Wei Lin und Shirui Pan. „CGMN: A Contrastive Graph Matching Network for Self-Supervised Graph Similarity Learning“. In Thirty-First International Joint Conference on Artificial Intelligence {IJCAI-22}. California: International Joint Conferences on Artificial Intelligence Organization, 2022. http://dx.doi.org/10.24963/ijcai.2022/292.
Der volle Inhalt der QuelleAnnotation:
Graph similarity learning refers to calculating the similarity score between two graphs, which is required in many realistic applications, such as visual tracking, graph classification, and collaborative filtering. As most of the existing graph neural networks yield effective graph representations of a single graph, little effort has been made for jointly learning two graph representations and calculating their similarity score. In addition, existing unsupervised graph similarity learning methods are mainly clustering-based, which ignores the valuable information embodied in graph pairs. To this end, we propose a contrastive graph matching network (CGMN) for self-supervised graph similarity learning in order to calculate the similarity between any two input graph objects. Specifically, we generate two augmented views for each graph in a pair respectively. Then, we employ two strategies, namely cross-view interaction and cross-graph interaction, for effective node representation learning. The former is resorted to strengthen the consistency of node representations in two views. The latter is utilized to identify node differences between different graphs. Finally, we transform node representations into graph-level representations via pooling operations for graph similarity computation. We have evaluated CGMN on eight real-world datasets, and the experiment results show that the proposed new approach is superior to the state-of-the-art methods in graph similarity learning downstream tasks.
5
Guan, Sheng, Hanchao Ma und Yinghui Wu. „RoboGNN: Robustifying Node Classification under Link Perturbation“. In Thirty-First International Joint Conference on Artificial Intelligence {IJCAI-22}. California: International Joint Conferences on Artificial Intelligence Organization, 2022. http://dx.doi.org/10.24963/ijcai.2022/420.
Der volle Inhalt der QuelleAnnotation:
Graph neural networks (GNNs) have emerged as powerful approaches for graph representation learning and node classification. Nevertheless, they can be vulnerable (sensitive) to link perturbations due to structural noise or adversarial attacks. This paper introduces RoboGNN, a novel framework that simultaneously robustifies an input classifier to a counterpart with certifiable robustness, and suggests desired graph representation with auxiliary links to ensure the robustness guarantee. (1) We introduce (p,θ)-robustness, which characterizes the robustness guarantee of a GNN-based classifier if its performance is insensitive for at least θ fraction of a targeted set of nodes under any perturbation of a set of vulnerable links up to a bounded size p. (2) We present a co-learning framework that interacts model learning with graph structural learning to robustify an input model M to a (p,θ)-robustness counterpart. The framework also outputs the desired graph structures that ensure the robustness. Using real-world benchmark graphs, we experimentally verify that roboGNN can effectively robustify representative GNNs with guaranteed robustness, and desirable gains on accuracy.
6
Ahmetaj, Shqiponja, Robert David, Magdalena Ortiz, Axel Polleres, Bojken Shehu und Mantas Šimkus. „Reasoning about Explanations for Non-validation in SHACL“. In 18th International Conference on Principles of Knowledge Representation and Reasoning {KR-2021}. California: International Joint Conferences on Artificial Intelligence Organization, 2021. http://dx.doi.org/10.24963/kr.2021/2.
Der volle Inhalt der QuelleAnnotation:
The Shapes Constraint Language (SHACL) is a recently standardized language for describing and validating constraints over RDF graphs. The SHACL specification describes the so-called validation reports, which are meant to explain to the users the outcome of validating an RDF graph against a collection of constraints. Specifically, explaining the reasons why the input graph does not satisfy the constraints is challenging. In fact, the current SHACL standard leaves it open on how such explanations can be provided to the users. In this paper, inspired by works on logic-based abduction and database repairs, we study the problem of explaining non-validation of SHACL constraints. In particular, in our framework non-validation is explained using the notion of a repair, i.e., a collection of additions and deletions whose application on an input graph results in a repaired graph that does satisfy the given SHACL constraints. We define a collection of decision problems for reasoning about explanations, possibly restricting to explanations that are minimal with respect to cardinality or set inclusion. We provide a detailed characterization of the computational complexity of those reasoning tasks, including the combined and the data complexity.
7
Li, Zuchao, Xingyi Guo, Letian Peng, Lefei Zhang und Hai Zhao. „iRe2f: Rethinking Effective Refinement in Language Structure Prediction via Efficient Iterative Retrospecting and Reasoning“. In Thirty-Second International Joint Conference on Artificial Intelligence {IJCAI-23}. California: International Joint Conferences on Artificial Intelligence Organization, 2023. http://dx.doi.org/10.24963/ijcai.2023/570.
Der volle Inhalt der QuelleAnnotation:
Refinement plays a critical role in language structure prediction, a process that deals with complex situations such as structural edge interdependencies. Since language structure prediction usually modeled as graph parsing, typical refinement methods involve taking an initial parsing graph as input and refining it using language input and other relevant information. Intuitively, a refinement component, i.e., refiner, should be lightweight and efficient, as it is only responsible for correcting faults in the initial graph. However, current refiners add a significant burden to the parsing process due to their reliance on time-consuming encoding-decoding procedure on the language input and graph. To make the refiner more practical for real-world applications, this paper proposes a lightweight but effective iterative refinement framework, iRe^2f, based on iterative retrospecting and reasoning without involving the re-encoding process on the graph. iRe^2f iteratively refine the parsing graph based on interaction between graph and sequence and efficiently learns the shortcut to update the sequence and graph representations in each iteration. The shortcut is calculated based on the graph representation in the latest iteration. iRe^2f reduces the number of refinement parameters by 90% compared to the previous smallest refiner. Experiments on a variety of language structure prediction tasks show that iRe^2f performs comparably or better than current state-of-the-art refiners, with a significant increase in efficiency.
8
Fan, Zhihao, Zhongyu Wei, Siyuan Wang, Ruize Wang, Zejun Li, Haijun Shan und Xuanjing Huang. „TCIC: Theme Concepts Learning Cross Language and Vision for Image Captioning“. In Thirtieth International Joint Conference on Artificial Intelligence {IJCAI-21}. California: International Joint Conferences on Artificial Intelligence Organization, 2021. http://dx.doi.org/10.24963/ijcai.2021/91.
Der volle Inhalt der QuelleAnnotation:
Existing research for image captioning usually represents an image using a scene graph with low-level facts (objects and relations) and fails to capture the high-level semantics. In this paper, we propose a Theme Concepts extended Image Captioning (TCIC) framework that incorporates theme concepts to represent high-level cross-modality semantics. In practice, we model theme concepts as memory vectors and propose Transformer with Theme Nodes (TTN) to incorporate those vectors for image captioning. Considering that theme concepts can be learned from both images and captions, we propose two settings for their representations learning based on TTN. On the vision side, TTN is configured to take both scene graph based features and theme concepts as input for visual representation learning. On the language side, TTN is configured to take both captions and theme concepts as input for text representation re-construction. Both settings aim to generate target captions with the same transformer-based decoder. During the training, we further align representations of theme concepts learned from images and corresponding captions to enforce the cross-modality learning. Experimental results on MS COCO show the effectiveness of our approach compared to some state-of-the-art models.
9
Sun, Tien-Lung, Chuan-Jun Su, Richard J. Mayer und Richard A. Wysk. „Shape Similarity Assessment of Mechanical Parts Based on Solid Models“. In ASME 1995 Design Engineering Technical Conferences collocated with the ASME 1995 15th International Computers in Engineering Conference and the ASME 1995 9th Annual Engineering Database Symposium. American Society of Mechanical Engineers, 1995. http://dx.doi.org/10.1115/detc1995-0234.
Der volle Inhalt der QuelleAnnotation:
Abstract Previous research toward part similarity assessment has been based on two kinds of symbolic part representation schemes, i.e., GT coding schemes and form feature based descriptions. The performance of these approaches is limited because the symbolic part representation is information incomplete and ambiguous. A solid model, on the other hand, is an unambiguous and information complete computer representation for 3D objects. In this work, the shape similarity between two polyhedra is formalized based on isomorphic subgraphs of graph representations extracted from their boundary models. A novel approach is proposed to match subgraphs. Input to the proposed methodology consists of two types of solid models, i.e., the boundary model and a CSG representation called the TO tree. The TO tree is used to derive a set of major sweep directions for a part. Using the matched major sweep directions as probes, a generate-and-test strategy is applied to find the matched subgraphs. A prototype system based on the proposed methodology has been implemented. The prototype system was programmed in AutoLisp© and is interfaced to the AME© solid modeler of AutoCAD©. Shape similarity comparison between various mechanical parts with intersecting features has been successfully accomplished by the system.
10
Miller, Michael G., James L. Mathieson, Joshua D. Summers und Gregory M. Mocko. „Representation: Structural Complexity of Assemblies to Create Neural Network Based Assembly Time Estimation Models“. In ASME 2012 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference. American Society of Mechanical Engineers, 2012. http://dx.doi.org/10.1115/detc2012-71337.
Der volle Inhalt der QuelleAnnotation:
Assembly time estimation is traditionally a time intensive manual process requiring detailed geometric and process information to be available to a human designer. As a result of these factors, assembly time estimation is rarely applied during early design iterations. This paper explores the possibility that the assembly time estimation process can be automated while reducing the level of design detail required. The approach presented here trains artificial neural networks (ANNs) to estimate the assembly times of vehicle sub-assemblies at various stages using properties of the connectivity graph at that point as input data. Effectiveness of estimation is evaluated based on the distribution of estimates provided by a population of ANNs trained on the same input data using varying initial conditions. Results suggest that the method presented here can complete the time estimation of an assembly process with +/− 15% error given an initial sample of manually estimated times for the given sub-assembly.