In this paper,we introduce a visual analytics approach aimed at helping machine learning experts analyze the hidden states of layers in recurrent neural networks.Our technique allows the user to interactively inspect ...In this paper,we introduce a visual analytics approach aimed at helping machine learning experts analyze the hidden states of layers in recurrent neural networks.Our technique allows the user to interactively inspect how hidden states store and process information throughout the feeding of an input sequence into the network.The technique can help answer questions,such as which parts of the input data have a higher impact on the prediction and how the model correlates each hidden state configuration with a certain output.Our visual analytics approach comprises several components:First,our input visualization shows the input sequence and how it relates to the output(using color coding).In addition,hidden states are visualized through a nonlinear projection into a 2-D visualization space using t-distributed stochastic neighbor embedding to understand the shape of the space of the hidden states.Trajectories are also employed to show the details of the evolution of the hidden state configurations.Finally,a time-multi-class heatmap matrix visualizes the evolution of the expected predictions for multi-class classifiers,and a histogram indicates the distances between the hidden states within the original space.The different visualizations are shown simultaneously in multiple views and support brushing-and-linking to facilitate the analysis of the classifications and debugging for misclassified input sequences.To demonstrate the capability of our approach,we discuss two typical use cases for long short-term memory models applied to two widely used natural language processing datasets.展开更多
Purpose.Persistent corneal epithelial defects(PED)pre-sent a very challenging problem to an terior segment sur-geons.Autologous serum tears had been demonstrated to be beneficial in the treatment of PED.T he current s...Purpose.Persistent corneal epithelial defects(PED)pre-sent a very challenging problem to an terior segment sur-geons.Autologous serum tears had been demonstrated to be beneficial in the treatment of PED.T he current study was conducted to review the local spectr um of indications and to examine the outcome of autologous se rum tear usage.Methods.All cases of PED treated with autologous serum tears at a tertiary referral centre f or the period August1999-July 2001were identified and r eviewed.Results.A total of 10eyes from 10patients were identified(5OD :5OS).The gender ratio was 7M :3F and the me an age was36.8(range 17-73)years old.The mean duration of PED before the usage of autologous se rum tears was 22.4±69.6days.Six eyes healed within 2w eeks,but two eyes failed to heal after 1month of tr eatment and two pa-tients defaulted follow-up.No adverse effects were ob-served with the addition of autoseru m tears.Conclusions.The results of the current study correlated well with previ-ous reported studies.Autologous se rum,tears may be considered as a valuable adjunct in t he management of re-calcitrant cases of PED.展开更多
Semi-Lagrangian texture advection(SLTA)enables efficient visualization of 2D and 3D unsteady flow.The major drawback of SLTA-based visualizations is numerical diffusion caused by iterative texture interpolation.We foc...Semi-Lagrangian texture advection(SLTA)enables efficient visualization of 2D and 3D unsteady flow.The major drawback of SLTA-based visualizations is numerical diffusion caused by iterative texture interpolation.We focus on reducing numerical diffusion in techniques that use textures sparsely populated by solid blobs,such as typically in dye advection.A ReLU-based model architecture is the foundation of our ML-based approach.Multiple model configurations are trained to learn a performant interpolation model that reduces numerical diffusion.Our evaluation investigates the models’ability to generalize concerning the flow and length of the advection process.The model with the best tradeoff between the computational effort to compute,quality of the result,and generality of application is found to be single-layer ReLU-based.This model is further analyzed and explained in-depth and improved using symmetry constraints.Additionally,a metamodel is fitted to predict single-layer ReLU model parameters for advection processes of any length.The metamodel removes the need for any prior training when applying our technique to a new scenario.Additionally,we show that our model is compatible with Back and Forth Error Compensation and Correction to improve the quality of the advection result further.We demonstrate that our model shows excellent diffusion reduction properties in typical examples of 3D steady and unsteady flow visualization.Finally,we utilize the strong diffusion reduction capabilities of our model to compute dye advection with exponential decay,a novel method that we introduce to visualize the extent and evolution of unsteadiness in both 2D and 3D unsteady flow.展开更多
Purpose Uveal melanoma(UM)is the most common intraocular malignancy in adults.Previous studies have examined the intra-tumoral heterogeneity.However,the spatial distribution of tumor cells within the tumor microenviro...Purpose Uveal melanoma(UM)is the most common intraocular malignancy in adults.Previous studies have examined the intra-tumoral heterogeneity.However,the spatial distribution of tumor cells within the tumor microenvironment and its relationship with tumor progression still remains largely unclear.Our study aimed to analyze the correlation between cell distribution patterns and the prognosis of UM.Methods In this paper,we performed spatial transcriptomics(ST)sequencing on two UM samples to describe the different cellular distribution patterns.Gene Ontology(GO)and Kyoto Encyclopedia of Genes,Genomes(KEGG)functional enrichment analysis,and protein-protein interaction(PPI)network were performed to define the biological function of each cluster.Differentially expressed genes(DEGs)and survival analysis based on datasets from The Cancer Genome Atlas(TCGA)and Gene Expression Omnibus(GEO)database further confirmed the correlation between cellular distribution and clinical prognosis.Results We found two different patterns of tumor cell distribution.The focal tumor cells have a distinct ribosome synthesis and rRNA pathway.In contrast,the subpopulation tented to distribute diffusely was related to fatty acids metabolism profile,presumably supporting tumor growth by providing energy.The scattered tumor cell cluster was associated with malignant biological behaviors and was involved in extensive cellular interactions,including COLLAGEN.Moreover,pseudo-time analysis showed that migration started from the basal region through cell differentiation.According to the TCGA and GEO database,genes expressed characteristically in the scattered tumor cell cluster were related to poor prognosis.Conclusions Our study drew the ST maps for UM for the first time.These findings revealed the distribution patterns of tumor cells associated with different biological functions and pointed towards specific tumor subpopulations with higher invasiveness as potential therapeutic targets.Together,our study displayed an overview of UM transcriptome and explored the intra-tumoral heterogeneity of UM at the spatial level.展开更多
Dynamic Mode Decomposition(DMD)is a data-driven and model-free decomposition technique.It is suitable for revealing spatio-temporal features of both numerically and experimentally acquired data.Conceptually,DMD perfor...Dynamic Mode Decomposition(DMD)is a data-driven and model-free decomposition technique.It is suitable for revealing spatio-temporal features of both numerically and experimentally acquired data.Conceptually,DMD performs a low-dimensional spectral decomposition of the data into the following components:the modes,called DMD modes,encode the spatial contribution of the decomposition,whereas the DMD amplitudes specify their impact.Each associated eigenvalue,referred to as DMD eigenvalue,characterizes the frequency and growth rate of the DMD mode.In this paper,we demonstrate how the components of DMD can be utilized to obtain temporal and spatial information from time-dependent flow fields.We begin with the theoretical background of DMD and its application to unsteady flow.Next,we examine the conventional process with DMD mathematically and put it in relationship to the discrete Fourier transform.Our analysis shows that the current use of DMD components has several drawbacks.To resolve these problems we adjust the components and provide new and meaningful insights into the decomposition:we show that our improved components capture the spatio-temporal patterns of the flow better.Moreover,we remove redundancies in the decomposition and clarify the interplay between components,allowing users to understand the impact of components.These new representations,which respect the spatio-temporal character of DMD,enable two clustering methods that segment the flow into physically relevant sections and can therefore be used for the selection of DMD components.With a number of typical examples,we demonstrate that the combination of these techniques allows new insights with DMD for unsteady flow.展开更多
Importance. Medical images are essential for modern medicine and an important research subject in visualization. However,medical experts are often not aware of the many advanced three-dimensional (3D) medical image vi...Importance. Medical images are essential for modern medicine and an important research subject in visualization. However,medical experts are often not aware of the many advanced three-dimensional (3D) medical image visualization techniques thatcould increase their capabilities in data analysis and assist the decision-making process for specific medical problems. Ourpaper provides a review of 3D visualization techniques for medical images, intending to bridge the gap between medicalexperts and visualization researchers. Highlights. Fundamental visualization techniques are revisited for various medicalimaging modalities, from computational tomography to diffusion tensor imaging, featuring techniques that enhance spatialperception, which is critical for medical practices. The state-of-the-art of medical visualization is reviewed based on aprocedure-oriented classification of medical problems for studies of individuals and populations. This paper summarizes freesoftware tools for different modalities of medical images designed for various purposes, including visualization, analysis, andsegmentation, and it provides respective Internet links. Conclusions. Visualization techniques are a useful tool for medicalexperts to tackle specific medical problems in their daily work. Our review provides a quick reference to such techniques giventhe medical problem and modalities of associated medical images. We summarize fundamental techniques and readily availablevisualization tools to help medical experts to better understand and utilize medical imaging data. This paper could contributeto the joint effort of the medical and visualization communities to advance precision medicine.展开更多
Processing and visualizing large scale volumetric and geometric datasets is mission critical in an increasing number of applications in academic research as well as in commercial enterprise. Often the datasets are, or...Processing and visualizing large scale volumetric and geometric datasets is mission critical in an increasing number of applications in academic research as well as in commercial enterprise. Often the datasets are, or can be processed to become, sparse.In this paper, we present Vox Link, a novel approach to render sparse volume data in a memory-efficient manner enabling interactive rendering on common, offthe-shelf graphics hardware. Our approach utilizes current GPU architectures for voxelizing, storing, and visualizing such datasets. It is based on the idea of perpixel linked lists(pp LL), an A-buffer implementation for order-independent transparency rendering. The method supports voxelization and rendering of dense semi-transparent geometry, sparse volume data, and implicit surface representations with a unified data structure. The proposed data structure also enables efficient simulation of global lighting effects such as reflection, refraction, and shadow ray evaluation.展开更多
Dimensionality reduction is often used to project time series data from multidimensional to two-dimensional space to generate visual representations of the temporal evolution.In this context,we address the problem of ...Dimensionality reduction is often used to project time series data from multidimensional to two-dimensional space to generate visual representations of the temporal evolution.In this context,we address the problem of multidimensional time series visualization by presenting a new method to show and handle projection errors introduced by dimensionality reduction techniques on multidimensional temporal data.For visualization,subsequent time instances are rendered as dots that are connected by lines or curves to indicate the temporal dependencies.However,inevitable projection artifacts may lead to poor visualization quality and misinterpretation of the temporal information.Wrongly projected data points,inaccurate variations in the distances between projected time instances,and intersections of connecting lines could lead to wrong assumptions about the original data.We adapt local and global quality metrics to measure the visual quality along the projected time series,and we introduce a model to assess the projection error at intersecting lines.These serve as a basis for our new uncertainty visualization techniques that use different visual encodings and interactions to indicate,communicate,and work with the visualization uncertainty from projection errors and artifacts along the timeline of data points,their connections,and intersections.Our approach is agnostic to the projection method and works for linear and non-linear dimensionality reduction methods alike.展开更多
We present angle-uniform parallel coordinates,a data-independent technique that deforms the image plane of parallel coordinates so that the angles of linear relationships between two variables are linearly mapped alon...We present angle-uniform parallel coordinates,a data-independent technique that deforms the image plane of parallel coordinates so that the angles of linear relationships between two variables are linearly mapped along the horizontal axis of the parallel coordinates plot.Despite being a common method for visualizing multidimensional data,parallel coordinates are ineffective for revealing positive correlations since the associated parallel coordinates points of such structures may be located at infinity in the image plane and the asymmetric encoding of negative and positive correlations may lead to unreliable estimations.To address this issue,we introduce a transformation that bounds all points horizontally using an angleuniform mapping and shrinks them vertically in a structure-preserving fashion;polygonal lines become smooth curves and a symmetric representation of data correlations is achieved.We further propose a combined subsampling and density visualization approach to reduce visual clutter caused by overdrawing.Our method enables accurate visual pattern interpretation of data correlations,and its data-independent nature makes it applicable to all multidimensional datasets.The usefulness of our method is demonstrated using examples of synthetic and real-world datasets.展开更多
We investigate task performance and reading characteristics for scatterplots(Cartesian coordinates)and parallel coordinates.In a controlled eye-tracking study,we asked 24 participants to assess the relative distance o...We investigate task performance and reading characteristics for scatterplots(Cartesian coordinates)and parallel coordinates.In a controlled eye-tracking study,we asked 24 participants to assess the relative distance of points in multidimensional space,depending on the diagram type(parallel coordinates or a horizontal collection of scatterplots),the number of data dimensions(2,4,6,or 8),and the relative distance between points(15%,20%,or 25%).For a given reference point and two target points,we instructed participants to choose the target point that was closer to the reference point in multidimensional space.We present a visual scanning model that describes different strategies to solve this retrieval task for both diagram types,and propose corresponding hypotheses that we test using task completion time,accuracy,and gaze positions as dependent variables.Our results show that scatterplots outperform parallel coordinates significantly in 2 dimensions,however,the task was solved more quickly and more accurately with parallel coordinates in 8 dimensions.The eye-tracking data further shows significant differences between Cartesian and parallel coordinates,as well as between different numbers of dimensions.For parallel coordinates,there is a clear trend toward shorter fixations and longer saccades with increasing number of dimensions.Using an area-of-interest(AOI)based approach,we identify different reading strategies for each diagram type:For parallel coordinates,the participants’gaze frequently jumped back and forth between pairs of axes,while axes were rarely focused on when viewing Cartesian coordinates.We further found that participants’attention is biased:toward the center of the whole plot for parallel coordinates and skewed to the center/left side for Cartesian coordinates.We anticipate that these results may support the design of more effective visualizations for multidimensional data.展开更多
基金Funded by the Deutsche Forschungsgemeinschaft(German Research Foundation),No.251654672—TRR 161(Project B01)Germany’s Excellence Strategy,No.EXC-2075—390740016.
文摘In this paper,we introduce a visual analytics approach aimed at helping machine learning experts analyze the hidden states of layers in recurrent neural networks.Our technique allows the user to interactively inspect how hidden states store and process information throughout the feeding of an input sequence into the network.The technique can help answer questions,such as which parts of the input data have a higher impact on the prediction and how the model correlates each hidden state configuration with a certain output.Our visual analytics approach comprises several components:First,our input visualization shows the input sequence and how it relates to the output(using color coding).In addition,hidden states are visualized through a nonlinear projection into a 2-D visualization space using t-distributed stochastic neighbor embedding to understand the shape of the space of the hidden states.Trajectories are also employed to show the details of the evolution of the hidden state configurations.Finally,a time-multi-class heatmap matrix visualizes the evolution of the expected predictions for multi-class classifiers,and a histogram indicates the distances between the hidden states within the original space.The different visualizations are shown simultaneously in multiple views and support brushing-and-linking to facilitate the analysis of the classifications and debugging for misclassified input sequences.To demonstrate the capability of our approach,we discuss two typical use cases for long short-term memory models applied to two widely used natural language processing datasets.
文摘Purpose.Persistent corneal epithelial defects(PED)pre-sent a very challenging problem to an terior segment sur-geons.Autologous serum tears had been demonstrated to be beneficial in the treatment of PED.T he current study was conducted to review the local spectr um of indications and to examine the outcome of autologous se rum tear usage.Methods.All cases of PED treated with autologous serum tears at a tertiary referral centre f or the period August1999-July 2001were identified and r eviewed.Results.A total of 10eyes from 10patients were identified(5OD :5OS).The gender ratio was 7M :3F and the me an age was36.8(range 17-73)years old.The mean duration of PED before the usage of autologous se rum tears was 22.4±69.6days.Six eyes healed within 2w eeks,but two eyes failed to heal after 1month of tr eatment and two pa-tients defaulted follow-up.No adverse effects were ob-served with the addition of autoseru m tears.Conclusions.The results of the current study correlated well with previ-ous reported studies.Autologous se rum,tears may be considered as a valuable adjunct in t he management of re-calcitrant cases of PED.
基金funded by the Deutsche Forschungsgemein-schaft(DFG,German Research Foundation)under Germany’s Ex-cellence Strategy-EXC-2075-390740016.
文摘Semi-Lagrangian texture advection(SLTA)enables efficient visualization of 2D and 3D unsteady flow.The major drawback of SLTA-based visualizations is numerical diffusion caused by iterative texture interpolation.We focus on reducing numerical diffusion in techniques that use textures sparsely populated by solid blobs,such as typically in dye advection.A ReLU-based model architecture is the foundation of our ML-based approach.Multiple model configurations are trained to learn a performant interpolation model that reduces numerical diffusion.Our evaluation investigates the models’ability to generalize concerning the flow and length of the advection process.The model with the best tradeoff between the computational effort to compute,quality of the result,and generality of application is found to be single-layer ReLU-based.This model is further analyzed and explained in-depth and improved using symmetry constraints.Additionally,a metamodel is fitted to predict single-layer ReLU model parameters for advection processes of any length.The metamodel removes the need for any prior training when applying our technique to a new scenario.Additionally,we show that our model is compatible with Back and Forth Error Compensation and Correction to improve the quality of the advection result further.We demonstrate that our model shows excellent diffusion reduction properties in typical examples of 3D steady and unsteady flow visualization.Finally,we utilize the strong diffusion reduction capabilities of our model to compute dye advection with exponential decay,a novel method that we introduce to visualize the extent and evolution of unsteadiness in both 2D and 3D unsteady flow.
基金funded by National Natural Science Foundation of China(82220108017,82141128,82101180)The Capital Health Research and Development of Special(2024-1-2052)+2 种基金Science&Technology Project of Beijing Municipal Science&Technology Commission(Z201100005520045)Sanming Project of Medicine in Shenzhen(No.SZSM202311018)The Fundation for Beijing Science&Technology Development of TCM(BJZYYB-2023-17).
文摘Purpose Uveal melanoma(UM)is the most common intraocular malignancy in adults.Previous studies have examined the intra-tumoral heterogeneity.However,the spatial distribution of tumor cells within the tumor microenvironment and its relationship with tumor progression still remains largely unclear.Our study aimed to analyze the correlation between cell distribution patterns and the prognosis of UM.Methods In this paper,we performed spatial transcriptomics(ST)sequencing on two UM samples to describe the different cellular distribution patterns.Gene Ontology(GO)and Kyoto Encyclopedia of Genes,Genomes(KEGG)functional enrichment analysis,and protein-protein interaction(PPI)network were performed to define the biological function of each cluster.Differentially expressed genes(DEGs)and survival analysis based on datasets from The Cancer Genome Atlas(TCGA)and Gene Expression Omnibus(GEO)database further confirmed the correlation between cellular distribution and clinical prognosis.Results We found two different patterns of tumor cell distribution.The focal tumor cells have a distinct ribosome synthesis and rRNA pathway.In contrast,the subpopulation tented to distribute diffusely was related to fatty acids metabolism profile,presumably supporting tumor growth by providing energy.The scattered tumor cell cluster was associated with malignant biological behaviors and was involved in extensive cellular interactions,including COLLAGEN.Moreover,pseudo-time analysis showed that migration started from the basal region through cell differentiation.According to the TCGA and GEO database,genes expressed characteristically in the scattered tumor cell cluster were related to poor prognosis.Conclusions Our study drew the ST maps for UM for the first time.These findings revealed the distribution patterns of tumor cells associated with different biological functions and pointed towards specific tumor subpopulations with higher invasiveness as potential therapeutic targets.Together,our study displayed an overview of UM transcriptome and explored the intra-tumoral heterogeneity of UM at the spatial level.
文摘Dynamic Mode Decomposition(DMD)is a data-driven and model-free decomposition technique.It is suitable for revealing spatio-temporal features of both numerically and experimentally acquired data.Conceptually,DMD performs a low-dimensional spectral decomposition of the data into the following components:the modes,called DMD modes,encode the spatial contribution of the decomposition,whereas the DMD amplitudes specify their impact.Each associated eigenvalue,referred to as DMD eigenvalue,characterizes the frequency and growth rate of the DMD mode.In this paper,we demonstrate how the components of DMD can be utilized to obtain temporal and spatial information from time-dependent flow fields.We begin with the theoretical background of DMD and its application to unsteady flow.Next,we examine the conventional process with DMD mathematically and put it in relationship to the discrete Fourier transform.Our analysis shows that the current use of DMD components has several drawbacks.To resolve these problems we adjust the components and provide new and meaningful insights into the decomposition:we show that our improved components capture the spatio-temporal patterns of the flow better.Moreover,we remove redundancies in the decomposition and clarify the interplay between components,allowing users to understand the impact of components.These new representations,which respect the spatio-temporal character of DMD,enable two clustering methods that segment the flow into physically relevant sections and can therefore be used for the selection of DMD components.With a number of typical examples,we demonstrate that the combination of these techniques allows new insights with DMD for unsteady flow.
基金the Data for Better Health Project of Peking University-Master Kong and by NIH(R01 EB031872).
文摘Importance. Medical images are essential for modern medicine and an important research subject in visualization. However,medical experts are often not aware of the many advanced three-dimensional (3D) medical image visualization techniques thatcould increase their capabilities in data analysis and assist the decision-making process for specific medical problems. Ourpaper provides a review of 3D visualization techniques for medical images, intending to bridge the gap between medicalexperts and visualization researchers. Highlights. Fundamental visualization techniques are revisited for various medicalimaging modalities, from computational tomography to diffusion tensor imaging, featuring techniques that enhance spatialperception, which is critical for medical practices. The state-of-the-art of medical visualization is reviewed based on aprocedure-oriented classification of medical problems for studies of individuals and populations. This paper summarizes freesoftware tools for different modalities of medical images designed for various purposes, including visualization, analysis, andsegmentation, and it provides respective Internet links. Conclusions. Visualization techniques are a useful tool for medicalexperts to tackle specific medical problems in their daily work. Our review provides a quick reference to such techniques giventhe medical problem and modalities of associated medical images. We summarize fundamental techniques and readily availablevisualization tools to help medical experts to better understand and utilize medical imaging data. This paper could contributeto the joint effort of the medical and visualization communities to advance precision medicine.
基金partially funded by Deutsche Forschungsgemeinschaft (DFG) as part of SFB 716 project D.3, the Excellence Center at Linkoping and Lund in Information Technology (ELLIIT), and the Swedish e-Science Research Centre (SeRC)
文摘Processing and visualizing large scale volumetric and geometric datasets is mission critical in an increasing number of applications in academic research as well as in commercial enterprise. Often the datasets are, or can be processed to become, sparse.In this paper, we present Vox Link, a novel approach to render sparse volume data in a memory-efficient manner enabling interactive rendering on common, offthe-shelf graphics hardware. Our approach utilizes current GPU architectures for voxelizing, storing, and visualizing such datasets. It is based on the idea of perpixel linked lists(pp LL), an A-buffer implementation for order-independent transparency rendering. The method supports voxelization and rendering of dense semi-transparent geometry, sparse volume data, and implicit surface representations with a unified data structure. The proposed data structure also enables efficient simulation of global lighting effects such as reflection, refraction, and shadow ray evaluation.
基金Deutsche Forschungsgemeinschaft(DFG,German Research Foundation)under Germany’s Excellence Strategy–EXC-2075–390740016.
文摘Dimensionality reduction is often used to project time series data from multidimensional to two-dimensional space to generate visual representations of the temporal evolution.In this context,we address the problem of multidimensional time series visualization by presenting a new method to show and handle projection errors introduced by dimensionality reduction techniques on multidimensional temporal data.For visualization,subsequent time instances are rendered as dots that are connected by lines or curves to indicate the temporal dependencies.However,inevitable projection artifacts may lead to poor visualization quality and misinterpretation of the temporal information.Wrongly projected data points,inaccurate variations in the distances between projected time instances,and intersections of connecting lines could lead to wrong assumptions about the original data.We adapt local and global quality metrics to measure the visual quality along the projected time series,and we introduce a model to assess the projection error at intersecting lines.These serve as a basis for our new uncertainty visualization techniques that use different visual encodings and interactions to indicate,communicate,and work with the visualization uncertainty from projection errors and artifacts along the timeline of data points,their connections,and intersections.Our approach is agnostic to the projection method and works for linear and non-linear dimensionality reduction methods alike.
基金support from the Data for Better Health Project of Peking University-Master Kong,YW from the National Natural Science Foundation of China(62132017)DW from the Deutsche Forschungsgemeinschaft(DFG)Project-ID 251654672-TRR 161.
文摘We present angle-uniform parallel coordinates,a data-independent technique that deforms the image plane of parallel coordinates so that the angles of linear relationships between two variables are linearly mapped along the horizontal axis of the parallel coordinates plot.Despite being a common method for visualizing multidimensional data,parallel coordinates are ineffective for revealing positive correlations since the associated parallel coordinates points of such structures may be located at infinity in the image plane and the asymmetric encoding of negative and positive correlations may lead to unreliable estimations.To address this issue,we introduce a transformation that bounds all points horizontally using an angleuniform mapping and shrinks them vertically in a structure-preserving fashion;polygonal lines become smooth curves and a symmetric representation of data correlations is achieved.We further propose a combined subsampling and density visualization approach to reduce visual clutter caused by overdrawing.Our method enables accurate visual pattern interpretation of data correlations,and its data-independent nature makes it applicable to all multidimensional datasets.The usefulness of our method is demonstrated using examples of synthetic and real-world datasets.
基金We would like to thank the Carl-Zeiss-Foundation(Carl-Zeiss-Stiftung)the German Research Foundation(DFG)for financial support within project B01 of SFB/Transregio 161.
文摘We investigate task performance and reading characteristics for scatterplots(Cartesian coordinates)and parallel coordinates.In a controlled eye-tracking study,we asked 24 participants to assess the relative distance of points in multidimensional space,depending on the diagram type(parallel coordinates or a horizontal collection of scatterplots),the number of data dimensions(2,4,6,or 8),and the relative distance between points(15%,20%,or 25%).For a given reference point and two target points,we instructed participants to choose the target point that was closer to the reference point in multidimensional space.We present a visual scanning model that describes different strategies to solve this retrieval task for both diagram types,and propose corresponding hypotheses that we test using task completion time,accuracy,and gaze positions as dependent variables.Our results show that scatterplots outperform parallel coordinates significantly in 2 dimensions,however,the task was solved more quickly and more accurately with parallel coordinates in 8 dimensions.The eye-tracking data further shows significant differences between Cartesian and parallel coordinates,as well as between different numbers of dimensions.For parallel coordinates,there is a clear trend toward shorter fixations and longer saccades with increasing number of dimensions.Using an area-of-interest(AOI)based approach,we identify different reading strategies for each diagram type:For parallel coordinates,the participants’gaze frequently jumped back and forth between pairs of axes,while axes were rarely focused on when viewing Cartesian coordinates.We further found that participants’attention is biased:toward the center of the whole plot for parallel coordinates and skewed to the center/left side for Cartesian coordinates.We anticipate that these results may support the design of more effective visualizations for multidimensional data.
