Distinguished Speaker: MM++: Post-Hoc Scale-Invariant Multilayer OOD Detection via Top-K Gated Feature Fusion

We introduce MM++ (Multilayer Mahalanobis++), a strictly post-hoc, and scale-invariant framework for Out-of-Distribution (OOD) detection. To address the trade-off between scale invariance and hierarchical expressivity, MM++ constructs a principled joint feature space. It first identifies discriminative intermediate layers by measuring entropy density drops, which mark the boundaries of sharp semantic compression. By fusing these selected layers with the terminal representation, the framework captures latent cross-layer correlations while mitigating early-layer noise. Crucially, a Ledoit-Wolf regularized tied covariance matrix stabilizes this unified space, enabling reliable distance estimation. Requiring no auxiliary OOD data, classifier fine-tuning, or architectural modifications, MM++ delivers robust performance across distinct architectures for both near- and far-OOD detection.

Bio:

Kyoung-Don Kang is a professor at the School of Computing at SUNY Binghamton. His research interests include AI for Cyber-Physical Systems and IoT, real-time data management, and real-time embedded systems. He has served on several related NSF projects as the lead and sole PI. His work has also been supported by industry, including Samsung. He received a PhD in Computer Science from the University of Virginia in 2003. Before that, he worked as a researcher and software engineer at the Agency for Defense Development in Korea between 1992 and 1998.