DFKI
RPTU Kaiserslautern–Landau
We select relevant samples for a replay buffer in continual learning using only the geometry of a pretrained feature space. CARES is among the state of the art replay buffer selectors while being 2.3 to 8.8 times faster than competitiors.
We speed up diffusion classifiers by utilizing a label hierarchy and pruning unrelated paths.
We present HyperCore, a lightweight adaptive coreset selection framework designed for noisy environments. HyperCore utilizes per class hypersphere models and adaptively selects pruning thresholds.
We introduce PRISM, a framework that disentangles architectural priors for dataset distillation, outperforming single-teacher setups.
We introduce SubZeroCore, a novel, training-free coreset selection method that integrates submodular coverage and density into a single, unified objective.