Full Professor, Northeastern University
2 papers at NeurIPS 2025
H-SPLID learns salient features via latent space decomposition and we provide theoretical guarantees w.r.t. input perturbations.