Early Recognition of Ball Catching Success in Clinical Trials with RNN-Based Predictive Classification

Lang, Jana; Giese, Martin A.; Synofzik, Matthis; Ilg, Winfried; Otte, Sebastian

doi:10.1007/978-3-030-86380-7_36

Cited by 3 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Catching: The Catching dataset [Lang et al, 2021] (provided by personal permission) contains multivariate two-dimensional movement trajectories of healthy and pathological ball catching trials over 60 time steps. At each time step, 20 features capture the catcher's arm position as well as the position of the ball.…”

Section: Datasetsmentioning

confidence: 99%

Generating Sparse Counterfactual Explanations For Multivariate Time Series

Lang¹,

Giese²,

Ilg³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Since neural networks play an increasingly important role in critical sectors, explaining network predictions has become a key research topic. Counterfactual explanations can help to understand why classifier models decide for particular class assignments and, moreover, how the respective input samples would have to be modified such that the class prediction changes. Previous approaches mainly focus on image and tabular data. In this work we propose SPARCE 1 , a generative adversarial network (GAN) architecture that generates SPARse Counterfactual Explanations for multivariate time series. Our approach provides a custom sparsity layer and regularizes the counterfactual loss function in terms of similarity, sparsity, and smoothness of trajectories. We evaluate our approach on real-world human motion datasets as well as a synthetic time series interpretability benchmark. Although we make significantly sparser modifications than other approaches, we achieve comparable or better performance on all metrics. Moreover, we demonstrate that our approach predominantly modifies salient time steps and features, leaving non-salient inputs untouched.

show abstract

Section: Datasetsmentioning

confidence: 99%

Generating Sparse Counterfactual Explanations For Multivariate Time Series

Lang¹,

Giese²,

Ilg³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…One of the most widely accepted is the plausibility. Current approaches leverage generative models [12] [13] or prototype losses [14] to guide the search away from outlier examples. However, this often drives counterfactuals to the center of the data manifold, neglecting the fact that the original input could not be near that center, or even that it could be out-of-sample, thus increasing the total amount of changes needed to reach a plausible counterfactual.…”

Section: Introductionmentioning

confidence: 99%

Sub-SpaCE: Subsequence-based Sparse Counterfactual Explanations for Time Series Classification Problems

Refoyo,

Luengo

2023

Preprint

View full text Add to dashboard Cite

The interpretation of existing machine learning models has become a critical task to facilitate the widespread adoption of AI across different domains, leading to the apparition of eXplainable AI (XAI). However, there exists an imbalance between methods tailored for specific data domains, with dedicated approaches for time series data having received limited attention until now. Moreover, current approaches often overlook the unique challenges present in time series data. In this paper, we introduce Subsequence-based Sparse Counterfactual Explanations (Sub-SpaCE), a novel method tailored for time series classification problems. Sub-SpaCE employs genetic algorithms with customized mutation and initialization processes, promoting changes in a small number of subsequences to generate highly sparse and plausible counterfactual explanations. Our empirical evaluations on various datasets demonstrate Sub-SpaCE's superior performance, achieving an optimal balance between sparsity and plausibility in counterfactual explanations for time series data.

show abstract