Search papers, labs, and topics across Lattice.
This paper introduces IMFACT, a model-agnostic framework that generates counterfactual explanations for time series classifiers by substituting selected Intrinsic Mode Functions (IMFs) derived from Empirical Mode Decomposition. By employing a Nearest Unlike Neighbour (NUN) approach, the method ensures that the temporal structure of oscillatory signals is preserved while effectively flipping the classifier's output to a target class. The results show that the variance-based IMF selection combined with a multi-NUN cycling strategy significantly improves reliability and plausibility over existing baseline techniques on benchmark datasets.
Counterfactual explanations for time series classifiers can now be generated without sacrificing temporal integrity, leading to more reliable and plausible outcomes.
Oscillatory signals, such as vibration, carry class-discriminative information in specific frequency bands; perturbing them in raw feature space for counterfactual analysis easily destroys their temporal structure and produces physically implausible results. In this work, we introduce IMFACT (IMF-based counterfACTuals), a model-agnostic framework for generating plausible counterfactual explanations for time series classifiers that operates in the decomposition space of Empirical Mode Decomposition. An input signal is split into Intrinsic Mode Functions (IMFs), and selected IMFs are progressively substituted with those of a Nearest Unlike Neighbour (NUN) until the classifier flips to the target class. We evaluate six IMF-selection strategies and a multi-NUN cycling extension on two UCR benchmarks (FaultDetectionA, FruitFlies). The variance-based strategy with three NUNs outperforms two prominent baseline techniques on reliability and plausibility metrics, while cycling across three NUNs yields the best proximity across both datasets.
