Interactive Graph Construction
for Graph-Based Semi-Supervised Learning

Changjian Chen1   Zhaowei Wang1   Jing Wu2   Xiting Wang3   Lan-Zhe Guo4   Yu-Feng Li4   Shixia Liu1

1Tsinghua University       2Cardiff University       3Microsoft Research Asia       4Nanjing University

Teaser Image
Teaser Image

DataLinker: (a) the Filtering panel helps focus on nodes and edges of interest; (b) the Label Change view shows label changes as an evolving river; (c) the Sample view displays samples as a combination of a scatterplot, a node-link diagram, and a bar chart; (d) the Action Trail records the modification history; (e) the Information view shows the images of selected samples and their nearest neighbors.

Abstract

Semi-supervised learning (SSL) provides a way to improve the performance of prediction models (e.g. , classifier) via the usage of unlabeled data. An effective and widely used method is to construct a graph that describes the relationship between labeled and unlabeled data. Practical experience indicates that graph quality significantly affects the model performance. In this paper, we present a visual analysis method that interactively constructs a high-quality graph for better model performance. In particular, we propose an interactive graph construction method based on the large margin principle. We have developed a river visualization and a hybrid visualization that combines a scatterplot, a node-link diagram, and a bar chart, to convey the label propagation of graph-based SSL. Based on the understanding of the propagation, a user can select regions of interest to inspect and modify the graph. We conducted two case studies to showcase how our method facilitates the exploitation of labeled and unlabeled data to improve model performance.

Video, Demo, and Source Code

Video: http://datalinker.thuvis.org/Video

Demo: http://datalinker.thuvis.org/Demo

Source Code: https://github.com/chencjgene/SSLVis