FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3195098128> ?p ?o ?g. }

• W3195098128 abstract "Class imbalance and noisy labels are the norm rather than the exception in many large-scale classification datasets. Nevertheless, most works in machine learning typically assume balanced and clean data. There have been some recent attempts to tackle, on one side, the problem of learning from noisy labels and, on the other side, learning from long-tailed data. Each group of methods make simplifying assumptions about the other. Due to this separation, the proposed solutions often underperform when both assumptions are violated. In this work, we present a simple two-stage approach based on recent advances in self-supervised learning to treat both challenges simultaneously. It consists of, first, task-agnostic self-supervised pre-training, followed by task-specific fine-tuning using an appropriate loss. Most significantly, we find that self-supervised learning approaches are effectively able to cope with severe class imbalance. In addition, the resulting learned representations are also remarkably robust to label noise, when fine-tuned with an imbalance- and noise-resistant loss function. We validate our claims with experiments on CIFAR-10 and CIFAR-100 augmented with synthetic imbalance and noise, as well as the large-scale inherently noisy Clothing-1M dataset." @default.
• W3195098128 created "2021-08-30" @default.
• W3195098128 creator A5065291945 @default.
• W3195098128 creator A5065739477 @default.
• W3195098128 creator A5080948360 @default.
• W3195098128 date "2021-08-25" @default.
• W3195098128 modified "2023-10-16" @default.
• W3195098128 title "Learning From Long-Tailed Data With Noisy Labels" @default.
• W3195098128 cites W1921293667 @default.
• W3195098128 cites W2016053056 @default.
• W3195098128 cites W204268067 @default.
• W3195098128 cites W2117539524 @default.
• W3195098128 cites W2148143831 @default.
• W3195098128 cites W2163605009 @default.
• W3195098128 cites W2194775991 @default.
• W3195098128 cites W2308045930 @default.
• W3195098128 cites W2326925005 @default.
• W3195098128 cites W2340690086 @default.
• W3195098128 cites W2743200750 @default.
• W3195098128 cites W2757910899 @default.
• W3195098128 cites W2799269579 @default.
• W3195098128 cites W2883725317 @default.
• W3195098128 cites W2941387380 @default.
• W3195098128 cites W2962742544 @default.
• W3195098128 cites W2962762068 @default.
• W3195098128 cites W2962843773 @default.
• W3195098128 cites W2963081269 @default.
• W3195098128 cites W2963351448 @default.
• W3195098128 cites W2963399829 @default.
• W3195098128 cites W2963420272 @default.
• W3195098128 cites W2963691377 @default.
• W3195098128 cites W2963735582 @default.
• W3195098128 cites W2964121744 @default.
• W3195098128 cites W2964292098 @default.
• W3195098128 cites W2970941190 @default.
• W3195098128 cites W2981873476 @default.
• W3195098128 cites W2995181141 @default.
• W3195098128 cites W2995197345 @default.
• W3195098128 cites W2996108195 @default.
• W3195098128 cites W3005731330 @default.
• W3195098128 cites W3034978746 @default.
• W3195098128 cites W3035524453 @default.
• W3195098128 cites W3095121901 @default.
• W3195098128 cites W3100859887 @default.
• W3195098128 cites W3101821705 @default.
• W3195098128 cites W3102583815 @default.
• W3195098128 cites W3103846556 @default.
• W3195098128 cites W3105402527 @default.
• W3195098128 cites W3105422445 @default.
• W3195098128 cites W3107668149 @default.
• W3195098128 cites W3112617038 @default.
• W3195098128 cites W3122855191 @default.
• W3195098128 cites W3134652006 @default.
• W3195098128 cites W3158225217 @default.
• W3195098128 cites W3175562011 @default.
• W3195098128 cites W85350352 @default.
• W3195098128 doi "https://doi.org/10.48550/arxiv.2108.11096" @default.
• W3195098128 hasPublicationYear "2021" @default.
• W3195098128 type Work @default.
• W3195098128 sameAs 3195098128 @default.
• W3195098128 citedByCount "2" @default.
• W3195098128 countsByYear W31950981282021 @default.
• W3195098128 countsByYear W31950981282023 @default.
• W3195098128 crossrefType "posted-content" @default.
• W3195098128 hasAuthorship W3195098128A5065291945 @default.
• W3195098128 hasAuthorship W3195098128A5065739477 @default.
• W3195098128 hasAuthorship W3195098128A5080948360 @default.
• W3195098128 hasBestOaLocation W31950981281 @default.
• W3195098128 hasConcept C115961682 @default.
• W3195098128 hasConcept C119857082 @default.
• W3195098128 hasConcept C121332964 @default.
• W3195098128 hasConcept C127413603 @default.
• W3195098128 hasConcept C136389625 @default.
• W3195098128 hasConcept C14036430 @default.
• W3195098128 hasConcept C153180895 @default.
• W3195098128 hasConcept C154945302 @default.
• W3195098128 hasConcept C160920958 @default.
• W3195098128 hasConcept C201995342 @default.
• W3195098128 hasConcept C2776145971 @default.
• W3195098128 hasConcept C2777212361 @default.
• W3195098128 hasConcept C2778755073 @default.
• W3195098128 hasConcept C2780451532 @default.
• W3195098128 hasConcept C2781170535 @default.
• W3195098128 hasConcept C41008148 @default.
• W3195098128 hasConcept C50644808 @default.
• W3195098128 hasConcept C62520636 @default.
• W3195098128 hasConcept C78458016 @default.
• W3195098128 hasConcept C86803240 @default.
• W3195098128 hasConcept C99498987 @default.
• W3195098128 hasConceptScore W3195098128C115961682 @default.
• W3195098128 hasConceptScore W3195098128C119857082 @default.
• W3195098128 hasConceptScore W3195098128C121332964 @default.
• W3195098128 hasConceptScore W3195098128C127413603 @default.
• W3195098128 hasConceptScore W3195098128C136389625 @default.
• W3195098128 hasConceptScore W3195098128C14036430 @default.
• W3195098128 hasConceptScore W3195098128C153180895 @default.
• W3195098128 hasConceptScore W3195098128C154945302 @default.
• W3195098128 hasConceptScore W3195098128C160920958 @default.
• W3195098128 hasConceptScore W3195098128C201995342 @default.
• W3195098128 hasConceptScore W3195098128C2776145971 @default.

• next