Validation, Robustness, and Accuracy of Perturbation-Based Sensitivity Analysis Methods for Time-Series Deep Learning Models
It addresses the reliability of interpretability methods for time-series AI models, which is crucial for domains like healthcare and finance, but is incremental as it benchmarks existing methods.
This work evaluated perturbation-based sensitivity analysis methods for time-series deep learning models, finding that different methods yield inconsistent outputs and rankings, and their alignment with ground truth varies, with specific metrics showing up to 30% deviation in accuracy.
This work undertakes studies to evaluate Interpretability Methods for Time-Series Deep Learning. Sensitivity analysis assesses how input changes affect the output, constituting a key component of interpretation. Among the post-hoc interpretation methods such as back-propagation, perturbation, and approximation, my work will investigate perturbation-based sensitivity Analysis methods on modern Transformer models to benchmark their performances. Specifically, my work answers three research questions: 1) Do different sensitivity analysis (SA) methods yield comparable outputs and attribute importance rankings? 2) Using the same sensitivity analysis method, do different Deep Learning (DL) models impact the output of the sensitivity analysis? 3) How well do the results from sensitivity analysis methods align with the ground truth?