We are seeking master's students with a strong background in deep learning, mathematical modeling, and cognitive science to join a follow-up research project on self-explainable deep learning models. In the realm of artificial intelligence, achieving high performance is only part of the challenge; ensuring that models are transparent and their decisions are interpretable is equally crucial. To address this, our recent work emphasizes the importance of defining clear explainability principles and incorporating them directly into the objective functions of learning algorithms.
By formulating these principles mathematically, we aim to create models that inherently provide explanations for their predictions. For example, utilizing advanced mathematical tools allows us to quantify aspects of explainability. This approach may not only enhance the interpretability of the models but also maintain, or even improve, their classification performance.
Participants in this project will have the opportunity to apply these methodologies to diverse image datasets. The research aims to push the boundaries of how AI systems can be made more transparent and aligned with human cognitive processes. Ideal candidates are those who combine rigorous analytical skills with creative thinking, eager to explore innovative solutions at the intersection of deep learning and explainable AI.
[1] Sokol, Kacper, and Peter Flach. "Explainability fact sheets: a framework for systematic assessment of explainable approaches." Proceedings of the 2020 conference on fairness, accountability, and transparency. 2020.
[2] Choi, Changkyu, et al. "DIB-X: Formulating Explainability Principles for a Self-Explainable Model Through Information Theoretic Learning." ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024.