Graph Layouts by t‐SNE

Kruiger, Johannes F.; Rauber, Paulo E.; Martins, Rafael Messias; Kerren, Andreas; Kobourov, Stephen G.; Telea, Alexandru

doi:10.1111/cgf.13187

Cited by 73 publications

(103 citation statements)

References 54 publications

(78 reference statements)

Supporting

Mentioning

103

Contrasting

Order By: Relevance

“…The RVS, BH, and RVS&BH layout algorithms could achieve better layout quality on some graphs with different parameters. For example, Figure shows the dwt_1005 and plat1919 graphs after 300 iterations of the RVS layout algorithm with alpha https://www.overleaf.com/project/5c0c13833b5200372464d136deca α d = 0 and no gravitational force, which appears very similar to layouts produced by other algorithms [OKB16, KRM*17]. However, the goal of this paper is not to tune parameters for specific graph types, but rather to show that RVS has comparable layout quality as BH while achieving substantially faster performance.…”

Section: Discussion and Future Workmentioning

confidence: 74%

“…HDE [HK02], ACE [KCH02], SSDE [ÇMIBR06], and Pivot MDS [BP06] are fast (e.g. HDE, SSDE, and Pivot MDS are linear time), but they have known issues with preserving neighborhoods [KRM*17] and producing good layouts for tree‐like and other non‐mesh‐like graphs [KRM*17,ÇMIBR06,HJ07]. These algorithms have not been widely adopted, unlike force‐directed algorithms that enjoy widespread adoption in many software packages.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Random Sampling O(n) Force‐calculation Algorithm for Graph Layouts

Gove¹

2019

Computer Graphics Forum

View full text Add to dashboard Cite

This paper proposes a linear‐time repulsive‐force‐calculation algorithm with sub‐linear auxiliary space requirements, achieving an asymptotic improvement over the Barnes‐Hut and Fast Multipole Method force‐calculation algorithms. The algorithm, named random vertex sampling (RVS), achieves its speed by updating a random sample of vertices at each iteration, each with a random sample of repulsive forces. This paper also proposes a combination algorithm that uses RVS to derive an initial layout and then applies Barnes‐Hut to refine the layout. An evaluation of RVS and the combination algorithm compares their speed and quality on 109 graphs against a Barnes‐Hut layout algorithm. The RVS algorithm performs up to 6.1 times faster on the tested graphs while maintaining comparable layout quality. The combination algorithm also performs faster than Barnes‐Hut, but produces layouts that are more symmetric than using RVS alone. Data and code: https://osf.io/nb7m8/

show abstract

Section: Discussion and Future Workmentioning

confidence: 74%

Section: Related Workmentioning

confidence: 99%

A Random Sampling O(n) Force‐calculation Algorithm for Graph Layouts

Gove¹

2019

Computer Graphics Forum

View full text Add to dashboard Cite

show abstract

“…One approach is edge bundling, which routes edges that are related according to some metric of similarity in close proximity [Hol06, HvW09] (see the recent survey by Llhuillier et al [LHT17] for details on edge bundling). The similarity metric is frequently based on the topology of the network [KRM*17], so that edges that have a similar region of origin and destination are bundled together. Edge bundling reduces clutter and makes it easier to detect connectivity patterns in the network [PHT15].…”

Section: Multivariate Network Visualization Typologymentioning

confidence: 99%

The State of the Art in Visualizing Multivariate Networks

Nobre

Meyer

Streit

et al. 2019

Computer Graphics Forum

101

View full text Add to dashboard Cite

Multivariate networks are made up of nodes and their relationships (links), but also data about those nodes and links as attributes. Most real‐world networks are associated with several attributes, and many analysis tasks depend on analyzing both, relationships and attributes. Visualization of multivariate networks, however, is challenging, especially when both the topology of the network and the attributes need to be considered concurrently. In this state‐of‐the‐art report, we analyze current practices and classify techniques along four axes: layouts, view operations, layout operations, and data operations. We also provide an analysis of tasks specific to multivariate networks and give recommendations for which technique to use in which scenario. Finally, we survey application areas and evaluation methodologies.

show abstract

“…t-SNE is a non-linear algorithm that captures the global, as well as the local structure of high-dimensional data [18]. t-SNE is one of the best methods for use in dimensionality reduction [21]. t-SNE projects the neighborhood structure of the data in a way that points that are similar will remain closer in the space [18].…”

Section: A Dimensionality Reductionmentioning

confidence: 99%

DNA Methylation Data to Predict Suicidal and Non-Suicidal Deaths: A Machine Learning Approach

Zahan

McQuillan

Osgood

2018

2018 IEEE International Conference on Healthcare Informatics (ICHI)

View full text Add to dashboard Cite

Suicide is one of the leading causes of death. Major Depressive Disorder (MDD) is one of the risk factors for committing suicide. Epigenetic data may help to distinguish suicidal and non-suicidal deaths. In particular, DNA methylation is a process involving a chemical modification on DNA which can change gene activity without changing the sequence. It has successfully been used for monitoring suicide progression and prediction. The objective of this study is to predict suicidal and non-suicidal deaths from DNA methylation data using a modern machine learning algorithm. We used support-vector machines to classify existing secondary data consisting of normalized values of methylated DNA probe intensities from tissues of two cortical brain regions (Brodmann Area 11 [BA11] and Brodmann Area 25 [BA25]) to distinguish 20 suicide cases (who died following a major depressive disorder) from 20 control cases (non-psychiatric sudden death). Prior to classification, we employed Principal component analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) to reduce the dimension of the data. In comparison to PCA, the modern data visualization method t-SNE performs better in dimensionality reduction. t-SNE accounts for the possible non-linear patterns in low-dimensional data. By contrast, PCA failed to group the cases according to different types of deaths. We applied four-fold cross-validation in which the resulting output from t-SNE was used as training data for the Support Vector Machine (SVM). For BA25, two-dimensional t-SNE better classified the suicidal vs. non-suicidal deaths compared to three-dimensional t-SNE. The Receiver Operating Characteristic Curve (ROC) for the classification exhibited an 80% Area Under the Curve (AUC) for BA25 data and 100% AUC for BA11. Despite the use of cross-validation, the nominally perfect prediction of suicidal deaths for BA11 data suggests possible over-fitting of the model. The study also may have suffered from 'spectrum bias' since the individuals were only studied from two extreme scenarios. This research constitutes a baseline study for classifying suicidal and non-suicidal deaths from DNA methylation data. Future studies with larger sample size, while possibly incorporating methylation data from living individuals, may reduce the bias and improve the accuracy of the results.

show abstract

Graph Layouts by t‐SNE

Cited by 73 publications

References 54 publications

A Random Sampling O(n) Force‐calculation Algorithm for Graph Layouts

A Random Sampling O(n) Force‐calculation Algorithm for Graph Layouts

The State of the Art in Visualizing Multivariate Networks

DNA Methylation Data to Predict Suicidal and Non-Suicidal Deaths: A Machine Learning Approach

Contact Info

Product

Resources

About