Matches in SemOpenAlex for { <https://semopenalex.org/work/W3176361985> ?p ?o ?g. }
- W3176361985 abstract "Training on class-imbalanced data usually results in biased models that tend to predict samples into the majority classes, which is a common and notorious problem. From the perspective of energy-based model, we demonstrate that the free energies of categories are aligned with the label distribution theoretically, thus the energies of different classes are expected to be close to each other when aiming for ``balanced'' performance. However, we discover a severe energy-bias phenomenon in the models trained on class-imbalanced dataset. To eliminate the bias, we propose a simple and effective method named Energy Aligning by merely adding the calculated shift scalars onto the output logits during inference, which does not require to (i) modify the network architectures, (ii) intervene the standard learning paradigm, (iii) perform two-stage training. The proposed algorithm is evaluated on two class imbalance-related tasks under various settings: class incremental learning and long-tailed recognition. Experimental results show that energy aligning can effectively alleviate class imbalance issue and outperform state-of-the-art methods on several benchmarks." @default.
- W3176361985 created "2021-07-05" @default.
- W3176361985 creator A5003388740 @default.
- W3176361985 creator A5034104790 @default.
- W3176361985 creator A5063253432 @default.
- W3176361985 creator A5071290194 @default.
- W3176361985 date "2021-06-07" @default.
- W3176361985 modified "2023-09-27" @default.
- W3176361985 title "Energy Aligning for Biased Models." @default.
- W3176361985 cites W104184427 @default.
- W3176361985 cites W1515020792 @default.
- W3176361985 cites W1522301498 @default.
- W3176361985 cites W1682403713 @default.
- W3176361985 cites W1821462560 @default.
- W3176361985 cites W1941659294 @default.
- W3176361985 cites W2014208555 @default.
- W3176361985 cites W2060277733 @default.
- W3176361985 cites W2117539524 @default.
- W3176361985 cites W2118978333 @default.
- W3176361985 cites W2119191234 @default.
- W3176361985 cites W2151084831 @default.
- W3176361985 cites W2162651021 @default.
- W3176361985 cites W2302255633 @default.
- W3176361985 cites W2339172597 @default.
- W3176361985 cites W2549139847 @default.
- W3176361985 cites W2737492962 @default.
- W3176361985 cites W2743473392 @default.
- W3176361985 cites W2753300133 @default.
- W3176361985 cites W2765407302 @default.
- W3176361985 cites W2767106145 @default.
- W3176361985 cites W2786498526 @default.
- W3176361985 cites W2795282075 @default.
- W3176361985 cites W2899063268 @default.
- W3176361985 cites W2899771611 @default.
- W3176361985 cites W2910580498 @default.
- W3176361985 cites W2921861056 @default.
- W3176361985 cites W2922656120 @default.
- W3176361985 cites W2925387673 @default.
- W3176361985 cites W2939117106 @default.
- W3176361985 cites W2939137134 @default.
- W3176361985 cites W2947380870 @default.
- W3176361985 cites W2949457404 @default.
- W3176361985 cites W2953106684 @default.
- W3176361985 cites W2954929116 @default.
- W3176361985 cites W2962775818 @default.
- W3176361985 cites W2962895364 @default.
- W3176361985 cites W2963341956 @default.
- W3176361985 cites W2963453204 @default.
- W3176361985 cites W2963588172 @default.
- W3176361985 cites W2963703618 @default.
- W3176361985 cites W2964278684 @default.
- W3176361985 cites W2970199615 @default.
- W3176361985 cites W2981864462 @default.
- W3176361985 cites W2995197345 @default.
- W3176361985 cites W3011400958 @default.
- W3176361985 cites W3021827438 @default.
- W3176361985 cites W3034601242 @default.
- W3176361985 cites W3034856281 @default.
- W3176361985 cites W3035524453 @default.
- W3176361985 cites W3092527263 @default.
- W3176361985 cites W3106701815 @default.
- W3176361985 cites W3118608800 @default.
- W3176361985 cites W3122855191 @default.
- W3176361985 hasPublicationYear "2021" @default.
- W3176361985 type Work @default.
- W3176361985 sameAs 3176361985 @default.
- W3176361985 citedByCount "0" @default.
- W3176361985 crossrefType "posted-content" @default.
- W3176361985 hasAuthorship W3176361985A5003388740 @default.
- W3176361985 hasAuthorship W3176361985A5034104790 @default.
- W3176361985 hasAuthorship W3176361985A5063253432 @default.
- W3176361985 hasAuthorship W3176361985A5071290194 @default.
- W3176361985 hasConcept C105795698 @default.
- W3176361985 hasConcept C111472728 @default.
- W3176361985 hasConcept C119857082 @default.
- W3176361985 hasConcept C12713177 @default.
- W3176361985 hasConcept C138885662 @default.
- W3176361985 hasConcept C154945302 @default.
- W3176361985 hasConcept C186370098 @default.
- W3176361985 hasConcept C2776214188 @default.
- W3176361985 hasConcept C2777212361 @default.
- W3176361985 hasConcept C2780586882 @default.
- W3176361985 hasConcept C33923547 @default.
- W3176361985 hasConcept C41008148 @default.
- W3176361985 hasConceptScore W3176361985C105795698 @default.
- W3176361985 hasConceptScore W3176361985C111472728 @default.
- W3176361985 hasConceptScore W3176361985C119857082 @default.
- W3176361985 hasConceptScore W3176361985C12713177 @default.
- W3176361985 hasConceptScore W3176361985C138885662 @default.
- W3176361985 hasConceptScore W3176361985C154945302 @default.
- W3176361985 hasConceptScore W3176361985C186370098 @default.
- W3176361985 hasConceptScore W3176361985C2776214188 @default.
- W3176361985 hasConceptScore W3176361985C2777212361 @default.
- W3176361985 hasConceptScore W3176361985C2780586882 @default.
- W3176361985 hasConceptScore W3176361985C33923547 @default.
- W3176361985 hasConceptScore W3176361985C41008148 @default.
- W3176361985 hasLocation W31763619851 @default.
- W3176361985 hasOpenAccess W3176361985 @default.
- W3176361985 hasPrimaryLocation W31763619851 @default.
- W3176361985 hasRelatedWork W129457532 @default.