Higher-order Networks of Diabetes Comorbidities: Disease Trajectories that Matter

Krieg, Steven J.; Robertson, Daniel H.; Pradhan, Meeta; Chawla, Nitesh V.

doi:10.1109/ichi48887.2020.9374390

Cited by 8 publications

(6 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Data sets We generated FONs and HONs with k = 2 (details in Appendix I) for four real-world data sets with known higher-order dependencies: flight itineraries for airline passengers in the United States (Air) [35], disease trajectories for type 2 diabetes patients in Indiana (T2D) [23], clickstreams of users playing the Wikispeedia game (Wiki) [43], and readership trajectories for a large online magazine (NY) [41]. As summarized in Table 1, the graphs were relatively diverse with respect to size, density, differences between G 1 and G 2 , and average homophily H (defined as the fraction of edges that connect nodes of the same class [51]).…”

Section: Methodsmentioning

confidence: 99%

“…In the present work, we couple GNNs with a specific family of graphs, higher-order networks (HONs), which encode sequential dependencies (i.e., statistical patterns that cannot be accounted for in a first-order Markov model) in a graph structure [45]. HONs have been applied to the study of user behaviors [11], citation networks [30], human mobility and navigation patterns [36,28], the spread of invasive species [45,32], anomaly detection [33], disease progression [23], and many more [21,28,35,25]. A traditional graph, which we call a first-order network (FON), represents a system by decomposing it into a set of pairwise edges, so the only way to infer polyadic interactions is via transitive paths over adjacent nodes.…”

Section: Introductionmentioning

confidence: 99%

“…To address this issue, Xu et al introduced a model that allows one or more higher-order nodes to represent a single entity within a single-layer graph structure, while pruning nodes and paths that do not exhibit higher-order dependencies [45]. This HON paradigm has proven fruitful in a number of supervised and unsupervised settings [33,31,32,23]. However, to our knowledge it has not yet been successfully used as input to a GNN.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Ensembles for Graphs with Higher-order Dependencies

Krieg¹,

Burgis²,

Soga³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Graph neural networks (GNNs) continue to achieve state-of-the-art performance on many graph learning tasks, but rely on the assumption that a given graph is a sufficient approximation of the true neighborhood structure. In the presence of higher-order sequential dependencies, we show that the tendency of traditional graph representations to underfit each node's neighborhood causes existing GNNs to generalize poorly. To address this, we propose a novel Deep Graph Ensemble (DGE), which captures neighborhood variance by training an ensemble of GNNs on different neighborhood subspaces of the same node within a higher-order network structure. We show that DGE consistently outperforms existing GNNs on semisupervised and supervised tasks on four real-world data sets with known higher-order dependencies, even under a similar parameter budget. We demonstrate that learning diverse and accurate base classifiers is central to DGE's success, and discuss the implications of these findings for future work on GNNs.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Ensembles for Graphs with Higher-order Dependencies

Krieg¹,

Burgis²,

Soga³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Liu et al [5] proposed reinforcement learning to learn and recommend sequential treatments including oral antidiabetic drugs and insulins to optimize the long-term patient outcomes of type 1 diabetes. Krieg et al [6] proposed to use high-order networks to the model complex relationship between diseases. The proposed method is used to predict the disease states and reproduce the disease trajectories for type 2 diabetes.…”

Section: Examples Of Predictive Modeling In Healthcare With Different...mentioning

confidence: 99%

Explainable Artificial Intelligence for Predictive Modeling in Healthcare

Yang

2022

J Healthc Inform Res

View full text Add to dashboard Cite

The principle behind artificial intelligence is mimicking human intelligence in the way that it can perform tasks, recognize patterns, or predict outcomes through learning from the acquired data of various sources. Artificial intelligence and machine learning algorithms have been widely used in autonomous driving, recommender systems in electronic commerce and social media, fintech, natural language understanding, and question answering systems. Artificial intelligence is also gradually changing the landscape of healthcare research 25). The rule-based approach that relied on the curation of medical knowledge and the construction of robust decision rules had drawn significant attention in diagnosing diseases and clinical decision support since half a century ago. In recent years, machine learning algorithms such as deep learning that can account for complex interactions between features is shown to be promising in predictive modeling in healthcare (Deo in Circulation 132:1920-1930. Although many of these artificial intelligence and machine learning algorithms can achieve remarkably high performance, it is often difficult to be completely adopted in practical clinical environments due to the lack of explainability in some of these algorithms. Explainable artificial intelligence (XAI) is emerging to assist in the communication of internal decisions, behavior, and actions to health care professionals. Through explaining the prediction outcomes, XAI gains the trust of the clinicians as they may learn how to apply the predictive modeling in practical situations instead of blindly following the predictions. There are still many scenarios to explore how to make XAI effective in clinical settings due to the complexity of medical knowledge.

show abstract

“…While the speci c assumptions about the type of higher-order structures included in those models di er, they have in common that they generalise network models towards representations that go beyond pairwise, dyadic interactions. Recent works in this area have used higher-order models for non-Markovian patterns in paths on networks to study random walks and di usion processes [15,24,29], detect communities and assess node centralities [8,21,24,28,36], analyse memory e ects in clinical time series data [13,18,20], generate node embeddings and network visualisations based on temporal network data [22,25,32], detect anomalies in time series data on networks [16,26], or assess the controllability of complex systems [37]. Moreover, recent works have shown the bene t of multiorder models that combine multiple higher-order models, e.g., for the generalisation of PageRank to time series data [27] or the prediction of paths [12].…”

Section: Introductionmentioning

confidence: 99%

Predicting Influential Higher-Order Patterns in Temporal Network Data

Gote¹,

Perri²,

Scholtes³

2021

Preprint

View full text Add to dashboard Cite

Networks are frequently used to model complex systems comprised of interacting elements. While links capture the topology of direct interactions, the true complexity of many systems originates from higher-order patterns in paths by which nodes can indirectly in uence each other. Path data, representing ordered sequences of consecutive direct interactions, can be used to model these patterns. However, to avoid over tting, such models should only consider those higher-order patterns for which the data provide su cient statistical evidence. On the other hand, we hypothesise that network models, which capture only direct interactions, under t higher-order patterns present in data. Consequently, both approaches are likely to misidentify in uential nodes in complex networks. We contribute to this issue by proposing eight centrality measures based on MOGen, a multi-order generative model that accounts for all paths up to a maximum distance but disregards paths at higher distances. We compare MOGen-based centralities to equivalent measures for network models and path data in a prediction experiment where we aim to identify in uential nodes in out-of-sample data. Our results show strong evidence supporting our hypothesis. MOGen consistently outperforms both the network model and path-based prediction. We further show that the performance di erence between MOGen and the path-based approach disappears if we have su cient observations, con rming that the error is due to over tting.

show abstract

Higher-order Networks of Diabetes Comorbidities: Disease Trajectories that Matter

Cited by 8 publications

References 27 publications

Deep Ensembles for Graphs with Higher-order Dependencies

Deep Ensembles for Graphs with Higher-order Dependencies

Explainable Artificial Intelligence for Predictive Modeling in Healthcare

Predicting Influential Higher-Order Patterns in Temporal Network Data

Contact Info

Product

Resources

About