9 papers
Hard Cases, Bad Labels: Testing Error Exposure and Error Location in Uncertainty Sampling Under Bounded Label Noise
John Myron Uy
Active learning can reduce labeling cost by selecting informative examples, but the most uncertain examples may also be the hardest to label correctly. This study tests whether unc…
Z-PEFT: Zero-shot Backdoor Detection in Parameter-Efficient Fine-Tuning via Canonical Spectral Signatures
Nicola Pitzalis, Donald Shenaj, Giacomo Cignoni +3
Parameter-Efficient Fine-tuned (PEFT) models are frequently downloaded from open repositories by practitioners. This widespread practice creates a significant attack surface, as ma…
Streaming Continual Learning for Unified Adaptive Intelligence in Dynamic Environments
Federico Giannini, Giacomo Ziffer, Andrea Cossu +1
Developing effective predictive models becomes challenging in dynamic environments that continuously produce data and constantly change. Continual Learning (CL) and Streaming Machi…
A Practical Guide to Streaming Continual Learning
Andrea Cossu, Federico Giannini, Giacomo Ziffer +5
Continual Learning (CL) and Streaming Machine Learning (SML) study the ability of agents to learn from a stream of non-stationary data. Despite sharing some similarities, they addr…
Learning and Transferring Physical Models through Derivatives
Alessandro Trenta, Andrea Cossu, Davide Bacciu
We propose Derivative Learning (DERL), a supervised approach that models physical systems by learning their partial derivatives. We also leverage DERL to build physical models incr…
CLA: Latent Alignment for Online Continual Self-Supervised Learning
Giacomo Cignoni, Andrea Cossu, Alexandra Gomez-Villa +2
Self-supervised learning (SSL) is able to build latent representations that generalize well to unseen data. However, only a few SSL techniques exist for the online CL setting, wher…