Magnostics: Image-Based Search of Interesting Matrix Views for Guided Network Exploration

Abstract

In this work we address the problem of retrieving potentially interesting matrix views to support the exploration of networks. We introduce Matrix Diagnostics (or Magnostics), following in spirit related approaches for rating and ranking other visualization techniques, such as Scagnostics for scatter plots. Our approach ranks matrix views according to the appearance of specific visual patterns, such as blocks and lines, indicating the existence of topological motifs in the data, such as clusters, bi-graphs, or central nodes. Magnostics can be used to analyze, query, or search for visually similar matrices in large collections, or to assess the quality of matrix reordering algorithms. While many feature descriptors for image analyzes exist, there is no evidence how they perform for detecting patterns in matrices. In order to make an informed choice of feature descriptors for matrix diagnostics, we evaluate 30 feature descriptors-27 existing ones and three new descriptors that we designed specifically for MAGNOSTICS-with respect to four criteria: pattern response, pattern variability, pattern sensibility, and pattern discrimination. We conclude with an informed set of six descriptors as most appropriate for Magnostics and demonstrate their application in two scenarios; exploring a large collection of matrices and analyzing temporal networks.

HiPiler: Visual Exploration of Large Genome Interaction Matrices with Interactive Small Multiples

“…The domain experts also expressed desire to pick and search patterns manually in the matrix view, for example, as a means to supplement results of a pattern detection algorithm. Yet, this will require image-based pattern detection algorithms similar to Magnostics [7] but on a much larger scale and in an interactive fashion. Some domain experts mentioned that integrating visualizations of other genomic features into snippets would further assist in finding correlations.…”

“…Another application example are networks with thousands of nodes, represented as adjacency matrices, and which are found in biology (e.g., gene regulatory or protein interaction networks), social application (e.g., Facebook), or computer science (e.g., server networks). ROIs in adjacency matrices can represent topological cliques and clusters, subgraphs, or specific graph motifs resulting in specific visual patterns in the matrix [7]. Adapting snippet exploration to networks requires an appropriate matrix ordering [6] to create visual patterns as well as pattern extraction methods specific to network.…”

FDive: Learning Relevance Models Using Pattern-based Similarity Measures

“…JPEG COEFFICIENT HISTOGRAM [42] MPEG7 EDGE HISTOGRAM [45] PHOG [9] HOUGH [29] PROFILE [4] Texture Other GABOR [42] BLOCKS [4] HARALICK [25] COMPACTNESS [45] LOCAL BINARY PATTERN [27] MAGNOSTICS [4] TAMURA [58] STATISTICAL NOISE [4] Feature Space (FS): A feature space describes the set of all feature vectors created by an individual feature descriptor. Additionally, a feature descriptor implies a vector space, called feature space.…”

“…The basis for all experiments is the Quick, Draw! dataset [32] 4 . We Figure 7: FDIVE learns to differentiate Electron Microscopy (EM) images containing synapses from images that do not.…”

Features Advances to Automatically Find Images for Application to Clinical Decision Support

“…Automatic illustration identification has been explored for illustrations in medical publications which may assist a clinician in determining the usefulness of a particular publication for patient monitoring and treatment [1] 2) Hough transform shape detection for region of interest determination and segmentation (has been used for lung images) [10]; 3) color analysis of stains for region of interest labeling (has been used for malaria cell images) [13]; 4) connecting the user and the database through a search engine with a feedback neural network architecture [14]; 5) query system modeling human interaction [15]; 6) Big Data use with query forms [16]; 7) use of image "key points" to identify salient parts of an image [17]; 7) combining image and text information for matrix similarity assessment [18]; 8) threedimensional image analysis [8]; 9) latent topic models for computing image similarity [19]; 10) statistical model-based image feature extraction using the wavelet domain and a Kullback divergence-based similarity measure for CBIR [21]; and 11) localized texture characterizations for CBIR for remote sensing applications [22].…”