Learning Predictive Substructures with Regularization for Network Data

Dang, Xuan Hong; You, Hongyuan; Bogdanov, Petko; Singh, Ambuj K.

doi:10.1109/icdm.2015.56

Cited by 10 publications

(15 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Both methods su↵er from limited quality and high running time due to the need to explore a large space of connected subgraphs. An alternative family of approaches for NSP were recently proposed following an optimization strategy [7,8]. DIPS [8] is the state-of-art approach, which introduces a two-stage solution to learn subgraphs: discriminative subspace learning followed by matrix approximation.…”

Section: Related Workmentioning

confidence: 99%

“…An alternative family of approaches for NSP were recently proposed following an optimization strategy [7,8]. DIPS [8] is the state-of-art approach, which introduces a two-stage solution to learn subgraphs: discriminative subspace learning followed by matrix approximation. This method avoids the search in the exponential space of candidate subgraphs, thus, addressing major drawbacks of NGF and MINDS.…”

Section: Related Workmentioning

confidence: 99%

“…Such approaches su↵er limited prediction quality and/or instability due to the local exploration of the exponential space of connected subgraphs. Other approaches enforce the design principles within optimization frameworks [8,7]. Due to the inherent complexity of simultaneous optimization of all three, these methods partition the principles in independent steps resulting in sub-par performance.…”

mentioning

confidence: 99%

“…An illustration of the above phenomenon is demonstrated in Fig. 1a by superimposing the subgraphs selected by (i) L1DSL (one of the methods proposed in this paper), (ii) the state-of-the-art optimization approach DIPS [8] and a ground truth (GT) subgraph GT (a) GT vs DSL vs DIPS injected in a Synthetic dataset. When constrained to select a fixed-size subgraph, our proposed method L1DSL recovers the connected GT subgraph, while DIPS recovers the GT only partially due to its two-step independent enforcement of the three design principles.…”

mentioning

confidence: 99%

See 3 more Smart Citations

DSL: Discriminative Subgraph Learning via Sparse Self-Representation

Zhang

Bogdanov

2019

Proceedings of the 2019 SIAM International Conference on Data Mining

Self Cite

View full text Add to dashboard Cite

The goal in network state prediction (NSP) is to classify the global label associated with features embedded in a graph. This graph structure encoding feature relationships is the key distinctive aspect of NSP compared to classical supervised learning. NSP arises in various applications: gene expression samples embedded in a protein-protein interaction (PPI) network, temporal snapshots of infrastructure or sensor networks, and fMRI coherence network samples from multiple subjects to name a few. Instances from these domains are typically "wide" (more features than samples), and thus, feature sub-selection is required for robust generalizable prediction. How to best employ the network structure in order to learn succinct connected subgraphs encompassing the most discriminative features becomes a central challenge. Prior work employs connected subgraph growth and sampling or graph smoothing within optimization frameworks, resulting in either large variance of quality or weak control over the connectivity of selected subgraphs.In this work we propose an optimization framework for discriminative subgraph learning (DSL), which simultaneously enforces (i) sparsity, (ii) connectivity and (iii) high discriminative power of the resulting subgraphs of features. Our optimization algorithm is a single-step solution for the NSP and associated feature selection problem. It is rooted in the rich literature on maximal-margin optimization, spectral graph methods and sparse subspace self-representation. DSL simultaneously ensures solution interpretability and superior predictive power (up to 16% improvement in challenging instances compared to baselines), with execution times up to an hour for large instances.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

DSL: Discriminative Subgraph Learning via Sparse Self-Representation

Zhang

Bogdanov

2019

Proceedings of the 2019 SIAM International Conference on Data Mining

Self Cite

View full text Add to dashboard Cite

show abstract

“…The analysis of brain data has recently attracted much attention from the data mining community with convincing results demonstrated in [4,17,28,31]. However, unlike most previous studies which focus on a small number of subjects and especially not for the temporal development of the disease, we analyze a large scale cohort of 180 subjects obtained from http://www.adni-info.org/, and evenly distributed into three global states: normal control (NC), mild cognitive impairment (MCI) and Alzheimer's disease (AD).…”

Section: Real World Datasetmentioning

confidence: 99%

Subnetwork Mining with Spatial and Temporal Smoothness

Dang

You

Singh

et al. 2017

Proceedings of the 2017 SIAM International Conference on Data Mining

Self Cite

View full text Add to dashboard Cite

In many real-world applications, data is represented in the form of networks with structures and attributes changing over time. The dynamic changes not only happen at nodes/edges, forming local subnetwork processes, but also eventually influence global states of networks. The need to understand what these local network processes are, how they evolve and consequently govern the progression of global network states has become increasingly important. In this paper, we explore these questions and develop a novel algorithm for mining a succinct set of subnetworks that are predictive and evolve along with the progression of global network states. Our algorithm is designed in the framework of logistic regression that fits a model for multi-states of network samples. Its objective function considers both the spatial network topology and temporal smooth transition between adjacent global network states, and we show that its global optimum solution can be achieved via steepest descent. Extensive experimental analysis on both synthetic and real world datasets demonstrates the effectiveness of our algorithm against competing methods, not only in the prediction accuracy but also in terms of domain relevance of the discovered subnetworks.

show abstract

Big-Graphs: Querying, Mining, and Beyond

Khan

Ranu

2017

Handbook of Big Data Technologies

View full text Add to dashboard Cite

Knowledge graphs (KGs) such as DBpedia, Freebase, YAGO, Wikidata, and NELL were constructed to store large-scale, real-world facts as ⟨subject, predicate, object⟩ triples -that can also be modeled as a graph, where a node (a subject or an object) represents an entity with attributes, and a directed edge (a predicate) is a relationship between two entities. Querying KGs is critical in web search, question answering (QA), semantic search, personal assistants, fact checking, and recommendation. While significant progress has been made on KG construction and curation, thanks to deep learning recently we have seen a surge of research on KG querying and QA. The objectives of our survey are two-fold. First, research on KG querying has been conducted by several communities, such as databases, data mining, semantic web, machine learning, information retrieval, and natural language processing (NLP), with different focus and terminologies; and also in diverse topics ranging from graph databases, query languages, join algorithms, graph patterns matching, to more sophisticated KG embedding and natural language questions (NLQs). We aim at uniting different interdisciplinary topics and concepts that have been developed for KG querying. Second, many recent advances on KG and query embedding, multimodal KG, and KG-QA come from deep learning, IR, NLP, and computer vision domains. We identify important challenges of KG querying that received less attention by graph databases, and by the DB community in general, e.g., incomplete KG, semantic matching, multimodal data, and NLQs. We conclude by discussing interesting opportunities for the data management community, for instance, KG as a unified data model and vector-based query processing.

show abstract

Learning Predictive Substructures with Regularization for Network Data

Cited by 10 publications

References 26 publications

DSL: Discriminative Subgraph Learning via Sparse Self-Representation

DSL: Discriminative Subgraph Learning via Sparse Self-Representation

Subnetwork Mining with Spatial and Temporal Smoothness

Big-Graphs: Querying, Mining, and Beyond

Contact Info

Product

Resources

About