OoDAnalyzer: Interactive Analysis of Out-of-Distribution Samples

TL;DR

OoDAnalyzer is a visual analysis tool that combines an improved ensemble OoD detection method with a novel grid layout algorithm to identify and explain out-of-distribution samples interactively, aiding model understanding.

Contribution

It introduces a new interactive visual analysis approach integrating an enhanced ensemble detection method and a novel kNN-based grid layout algorithm for OoD sample analysis.

Findings

Effective identification of OoD samples demonstrated on multiple datasets.

The proposed grid layout algorithm is faster and nearly optimal.

Quantitative evaluations confirm the usefulness of OoDAnalyzer.

Abstract

One major cause of performance degradation in predictive models is that the test samples are not well covered by the training data. Such not well-represented samples are called OoD samples. In this paper, we propose OoDAnalyzer, a visual analysis approach for interactively identifying OoD samples and explaining them in context. Our approach integrates an ensemble OoD detection method and a grid-based visualization. The detection method is improved from deep ensembles by combining more features with algorithms in the same family. To better analyze and understand the OoD samples in context, we have developed a novel kNN-based grid layout algorithm motivated by Hall's theorem. The algorithm approximates the optimal layout and has $O(kN^2)$ time complexity, faster than the grid layout algorithm with overall best performance but $O(N^3)$ time complexity. Quantitative evaluation and case studies were performed on several datasets to demonstrate the effectiveness and usefulness of OoDAnalyzer.