Self-Paced Learning with Diversity

By Lu Jiang, Deyu Meng, Shoou-I Yu, Zhenzhong Lan, Shiguang Shan, Alexander G. Hauptmann. Neural Information Processing Systems, 2014

398 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/44606e1209a47d1fcf88b90e306db9e4b84fa2c5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Exploit the Unknown Gradually: One-Shot Video-Based Person Re-identification by Stepwise Learning (2018, background): “The self-paced paradigm is theoretically analyzed in [13, 23].”
Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting (2019, methodology): “Typical methods include self-paced learning(SPL) [26], iterative reweighting [27, 17] and multiple variants [28, 29, 30].”
Adaptive SVM+: Learning with Privileged Information for Domain Adaptation (2017, background): “Privileged information can appear in the form of additional features available only at training time [25], in the form of a curriculum learning strategy [4, 13] (i.”
A Survey on Curriculum Learning (2021, background): “(iii) SPL: if reliable prior knowledge or assumption is available, embedding it into the SPL objective always help [22], [41], [42], [134].”
MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels (2017, methodology): “For example, Jiang et al. (2014) introduced a curriculum of using easy and diverse samples.”
A Self-Paced Multiple-Instance Learning Framework for Co-Saliency Detection (2015, unknown): “To this end, we make the earliest effort to introduce multi-instance learning (MIL) [8] to co-saliency detection for jointly exploring the contrast between co-salient objects and their contexts, as well as the consistency of the co-salient objects within multiple images.”
Self-Paced Curriculum Learning (2015, methodology): “A more general regularizer consists of both ‖v‖1 and ‖v‖2,1 (Jiang et al. 2014b).”
Adaptive Smoothed Online Multi-Task Learning (2016, background): “Optimization (2) shares the same underlying insight with Self-Paced Learning (SPL) [14, 15] where the algorithm automatically learns the weights over data points during training.”
Training Curricula for Open Domain Answer Re-Ranking (2020, background): “g, our weighting approach may be beneficial for ensuring that ‘easy’ samples are ranked effectively. Another possible direction for future work could explore the use of self-paced learning techniques [19, 20], allowing the model to learn which training samples characteristics make a samples easy or ”
Co-Saliency Detection via a Self-Paced Multiple-Instance Learning Framework (2017, background): “…regularizer [18], this paper ameliorates the sample diversity term as a convex negative l 0:5;1 -norm, which not only more complies with the original SPL axiomic definition, but also naturally leads to a more rational real-valued sample weighting scheme rather than the binary one as in [18].”
A Competence-aware Curriculum for Visual Concepts Learning via Question Answering (2020, methodology): “Self-paced learning [33,27,28,51] quantifies the sample hardness by the training loss and formulates curriculum learning as an optimization problem by jointly modeling the sample selection and the learning objective.”
Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels (2019, background): “The regularization term G determines a curriculum (Jiang et al., 2015; 2014; Fan et al., 2017) or equivalently a weighting scheme to compute the latent weight v i to each example.”
What Objective Does Self-paced Learning Indeed Optimize? (2015, background): “5,1 norm on v, as utilized in [19] and [21], respectively.”
Suppressing Uncertainties for Large-Scale Facial Expression Recognition (2020, background): “MentorNet [20] and other self-paced learning methods [19, 30] solve this problem by alternating minimization, i.e. optimize one at a time while the other is held ﬁxed.”
Learning What Data to Learn (2017, methodology): “In our implementation, to improve the robustness of SPL, following the widely used trick in common SPL implementation (Jiang et al., 2014b), we ﬁlter training data using its loss rank in one mini-batch, rather than the absolute loss value 3 .”
Curriculum Loss: Robust Learning and Generalization against Label Corruption (2019, methodology): “Jiang et al. (Jiang et al., 2014) propose a self-paced learning method to select samples with diversity.”
Easy Questions First? A Case Study on Curriculum Learning for Question Answering (2016, methodology): “The techniques have been shown to be useful in some computer vision tasks (Lee and Grauman, 2011; Kumar et al., 2011; Tang et al., 2012; Supancic and Ramanan, 2013; Jiang et al., 2014a).”
Learning the Curriculum with Bayesian Optimization for Task-Specific Word Representation Learning (2016, background): “Diversity has been shown effective in related research on curriculum learning in language modeling, vision, and multimedia analysis (Bengio et al., 2009; Jiang et al., 2014).”
MentorNet: Regularizing Very Deep Neural Networks on Corrupted Labels (2017, background): “find better minima in a variety of computer vision problems [38, 26, 7, 16, 25, 44], face detection [26], object detection [7], video detection [16].”