Projections as visual aids for classification system design

Abstract

Dimensionality reduction is a compelling alternative for high-dimensional data visualization. This method provides insight into high-dimensional feature spaces by mapping relationships between observations (high-dimensional vectors) to low (two or three) dimensional spaces. These low-dimensional representations support tasks such as outlier and group detection based on direct visualization. Supervised learning, a subfield of machine learning, is also concerned with observations. A key task in supervised learning consists of assigning class labels to observations based on generalization from previous experience. Effective development of such classification systems depends on many choices, including features descriptors, learning algorithms, and hyperparameters. These choices are not trivial, and there is no simple recipe to improve classification systems that perform poorly. In this context, we first propose the use of visual representations based on dimensionality reduction (projections) for predictive feedback on classification efficacy. Second, we propose a projection-based visual analytics methodology, and supportive tooling, that can be used to improve classification systems through feature selection. We evaluate our proposal through experiments involving four datasets and three representative learning algorithms.

Constructing and Visualizing High-Quality Classifier Decision Boundary Maps

“…Simply put, we do not know what happens in the blank space between the scatterplot points. In particular, the decision boundaries ∂DZ(c) of the classifier are not explicitly visualized, leaving the user to guessing their actual position [4].…”

“…As such methods can become very complex, as in the case of deep learning (DL) methods, practitioners and users have challenges in understanding, customizing and trusting them [1,2]. To alleviate this, recent work has focused on visually explaining how ML techniques learn and take their decisions [3][4][5][6].…”

Multidimensional Data Exploration by Explicitly Controlled Animation

“…To do this, one way is to engineer discriminating features using the raw 17 ones present in the data. Projections can help us in determining how good the engineered features are: if we find a projection where same-class points are well separated into clusters then the features that the projection has used as input are a good start for building a good classifier [79]. Figure 6 shows our method applied on this dataset, with the right image showing the t-SNE projection.…”

Visual Analytics for Machine Learning: Computing and Leveraging Decision Boundary Maps

“…Finally, plotting the actual misclassi ed points atop of such a projection allows us to reason about which data attributes these have and how these may have caused problems to the classi er. These, and other, scenarios have been recently examined in recent literature [114][115][116]. In particular, Rauber et al [115] show that projections can be used as good predictors for the ease of constructing a good classi er from a given training set.…”

“…These, and other, scenarios have been recently examined in recent literature [114][115][116]. In particular, Rauber et al [115] show that projections can be used as good predictors for the ease of constructing a good classi er from a given training set. The idea is further developed in [13], who show how projections can be used to improve an existing classi er by semi-supervised training.…”