Towards plausibility in time series counterfactual explanations

Marcin Kostrzewa, Krzysztof Galus, Maciej Zięba · 2026

Read Towards plausibility in time series counterfactual explanations by Marcin Kostrzewa, Krzysztof Galus, Maciej Zięba (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-08349.txt.

Description

Kostrzewa and colleagues present a new method for making plausible counterfactuals for time series sorting. The approach performs gradient-based optimization straight in the input space. To enforce realism, the method joins soft-DTW matching with k-nearest neighbors from the target class, pushing new ones to adopt a lifelike time-based structure. The overall goal is a multi-faceted loss function weighing key traits.

Catalog record

CREATORMarcin Kostrzewa, Krzysztof Galus, Maciej Zięba
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29139
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.08349

Browse the full library · About the Tower of Babel