The School of Data Science and Analytics is pleased to announce the dissertation defense of Mr. Sanad Biswas on Monday, July 14th at 12:00 p.m. Please join us in person in Atrium J2225 or via Teams at this link.

Mr. Biswas’s dissertation is entitled “Feature Significance and Explainability in Neural Network.”

His committee is: Dr. Sherry Ni (chair), Dr. Linh Le, Dr. Jonathan Boardman, and Dr. Xinyan Zhang.

_________________________

Abstract:

As deep neural networks grow increasingly powerful, concerns about their opacity and interpretability escalate, hindering their trustworthiness in high-stakes scenarios. eXplainable AI (XAI) methods have emerged to enhance transparency and accountability in neural networks, emphasizing interpretability through global feature significance, which explains model behavior across entire datasets, and local feature attribution, which explains individual predictions. A critical yet overlooked aspect is the selection of appropriate baselines for attribution methods. This dissertation addresses these dimensions by proposing rigorous methodologies: (1) a permutation-based testing framework for global feature significance, uniquely permuting the target variable to robustly handle nonlinear relationships and multicollinearity without restrictive assumptions; (2) statistical hypothesis tests and confidence intervals for local feature attribution methods, including Integrated Gradients, DeepLIFT, SHAP, and LIME, providing robust validation of individual feature contributions; and (3) a generative contrastive baseline approach that constructs minimally perturbed counterfactuals, offering actionable, realistic explanations within the data manifold. Together, these methodologies significantly advance XAI, integrating statistical rigor with practical applicability to promote transparent, accountable, and ethically responsible neural network models.