AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Zero-shot AutoML with Pretrained Models

Ekrem Öztürk, Fabio Ferreira, Hadi Jomaa, Lars Schmidt-Thieme, Josif Grabocka, Frank Hutter

ICML 2022 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Robust fine-tuning of zero-shot models · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Select a pipeline using geometry-aware ZAP-HPO rather than challenge-winning systemsZAP pipeline selection · ZAP benchmarkRank (ALC): 2.07±0.05 vs. 2.53±0.06 (DeepWisdom), 2.64±0.04 (DeepBlueAI), 2.76±0.03 (PASA NJU)Improved
Select a pipeline using geometry-aware ZAP-HPO rather than challenge-winning systemsZAP pipeline selection · ZAP benchmarkRank (Normalized AUC): 2.08±0.03 vs. 2.63±0.03 (DeepWisdom), 2.73±0.03 (DeepBlueAI), 2.56±0.03 (PASA NJU)Improved
Use ZAP-HPO instead of challenge-winning systemsPipeline selection on AutoDL benchmark · AutoDL benchmarkRank (ALC): 2.16±0.15 vs. 2.46±0.13 (DeepWisdom), 2.76±0.08 (DeepBlueAI), 2.62±0.11 (PASA NJU)Improved
Train ZAP-HPO with only 75% of cost-matrix entries (ablation)ZAP-HPO · ZAP benchmarkRank (ALC): 2.09±0.06No clear effect
Train ZAP-HPO with only 50% of cost-matrix entries (ablation)ZAP-HPO · ZAP benchmarkRank (ALC): 2.09±0.05No clear effect
Train ZAP-HPO with only 25% of cost-matrix entries (ablation)ZAP-HPO · ZAP benchmarkRank (ALC): 2.10±0.04No clear effect
Replace the pairwise ranking objective with least-squares regression (ablation)ZAP-HPO surrogate · ZAP benchmarkPipeline identification performanceHurt
Replace the pairwise ranking objective with a triplet margin objective (ablation)ZAP-HPO surrogate · ZAP benchmarkPipeline selection performanceHurt
Use ZAP-AS instead of ZAP-HPO for out-of-distribution selectionZAP pipeline selection · AutoDL benchmarkRank (ALC): 1.81 (± 0.3) vs. 2.16 (± 0.15)Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy