Exploring Multi‐dimensional Data via Subset Embedding

Abstract

Multi‐dimensional data exploration is a classic research topic in visualization. Most existing approaches are designed for identifying record patterns in dimensional space or subspace. In this paper, we propose a visual analytics approach to exploring subset patterns. The core of the approach is a subset embedding network (SEN) that represents a group of subsets as uniformly‐formatted embeddings. We implement the SEN as multiple subnets with separate loss functions. The design enables to handle arbitrary subsets and capture the similarity of subsets on single features, thus achieving accurate pattern exploration, which in most cases is searching for subsets having similar values on few features. Moreover, each subnet is a fully‐connected neural network with one hidden layer. The simple structure brings high training efficiency. We integrate the SEN into a visualization system that achieves a 3‐step workflow. Specifically, analysts (1) partition the given dataset into subsets, (2) select portions in a projected latent space created using the SEN, and (3) determine the existence of patterns within selected subsets. Generally, the system combines visualizations, interactions, automatic methods, and quantitative measures to balance the exploration flexibility and operation efficiency, and improve the interpretability and faithfulness of the identified patterns. Case studies and quantitative experiments on multiple open datasets demonstrate the general applicability and effectiveness of our approach.

VA + Embeddings STAR: A State‐of‐the‐Art Report on the Use of Embeddings in Visual Analytics

“…Visualization for Embedding Interpretation Embeddings are ubiquitous in machine learning. Beyond designing a panel for embedding analysis in a larger system of ML model interpretation as mentioned in Section 5.1.1, there exist many visual analytics papers dedicated to interpreting user‐defined embeddings [STN * 16] or exploring user‐defined multi‐dimensional data via embeddings [XTL * 21]. Smilkov et al [STN * 16] summarized three high‐level tasks to facilitate the interpretation of embeddings: (1) exploring local neighborhoods, (2) viewing global geometry and finding clusters, and (3) finding semantically meaningful directions for a certain concept set.…”

Improved Evaluation and Generation of Grid Layouts using Distance Preservation Quality and Linear Assignment Sorting

“…Since the run time strongly depends on the hardware and implementation quality, the numbers given in this section only serve as comparative values. In the previous section, DPQ 16 has shown high correlation with user preferences and performance, we therefore use it when comparing algorithms in terms of their achieved "quality" and the run time required to generate the sorted arrangement.…”

“…Numerous techniques (Principal Component Analysis (PCA) [9], Multidimensional Scaling (MDS) [10], Locally Linear Embedding (LLE) [11], Isomap [12], and others) are described in in [13]. Other methods that work very well are t-Distributed Stochastic Neighborhood Embedding (t-SNE) [14], Uniform Manifold Approximation and Projection (UMAP) [15] and Subset Embedding Networks [16].…”

Improved Evaluation and Generation Of Grid Layouts Using Distance Preservation Quality and Linear Assignment Sorting

“…Numerous techniques ( principal component analysis (PCA) [Pea01], multi‐dimensional scaling (MDS) [Sam69], locally linear embedding (LLE) [RS00], Isomap [TdSL00] and others) are described in Sarveniazi [Sar14]. Other methods that work very well are t‐distributed stochastic neighbourhood embedding (t‐SNE) [vdMH08], uniform manifold approximation and projection (UMAP) [MH18] and subset embedding networks [XTL*21].…”