Curriculum learning
The DOI matches its Crossref record; 35 of the 40 references it cites were found in a registry or an index.
Abstract
Humans and animals learn much better when the examples are not randomly presented but organized in a meaningful order which illustrates gradually more concepts, and gradually more complex ones. Here, we formalize such training strategies in the context of machine learning, and call them "curriculum learning". In the context of recent research studying the difficulty of training in the presence of non-convex training criteria (for deep deterministic and stochastic neural networks), we explore curriculum learning in various set-ups. The experiments show that significant improvements in generalization can be achieved. We hypothesize that curriculum learning has both an effect on the speed of convergence of the training process to a minimum and, in the case of non-convex criteria, on the quality of the local minima obtained: curriculum learning can be seen as a particular form of continuation method (a general strategy for global optimization of non-convex functions).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Moreover, the lack of acoustical signals could even account for spatial disorientation 91,92.”
From The impact of multisensory integration and perceptual load in virtual reality settings on performance, workload and presence (2021) · cites this paper as [92]
In the citing paper
“This suggests that curriculum-style learning (Bengio et al., 2009), where simpler or archetypal problems are presented first to build a repository of valid heuristics, may potentially bootstrap performance.”
From Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory (2025) · cites this paper as Bengio et al., 2009
Citations
Cites
Reducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
A fast learning algorithm for deep belief nets
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006) · Neural Computation · doi:10.1162/neco.2006.18.7.1527 · Cited by 16,336
Learning Deep Architectures for AI
Bengio, Y. (2009) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000006 · Cited by 9,239
Self-citationExtracting and composing robust features with denoising autoencoders
Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P.-A. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390294 · Cited by 7,436
Self-citationA unified architecture for natural language processing: deep neural networks with multitask learning
Collobert, R., & Weston, J. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390177 · Cited by 5,058 · cited 6×
“In Collobert and Weston (2008), the main objective is to learn an embedding for the words as a side-e.ect of learning to compute this score.”
BackgroundMethodInfluentialSelf-citationLearning Deep Architectures for AI
Bengio, Y. (2009) · now publishers, Inc. eBooks · doi:10.1561/9781601982957 · Cited by 4,941
Self-citationGreedy layer-wise training of deep networks
Bengio, Y., Lamblin, P., Popovici, D., & Larochelle, H. (2007) · Advances in Neural Information Processing Systems 19 · doi:10.7551/mitpress/7503.003.0024 · Cited by 4,841
Self-citationA Neural Probabilistic Language Model
Bengio, Y., Ducharme, R., Vincent, P., Jauvin, C., Jaz, K., Hofmann, T., Poggio, T., & Shawe-Taylor, J. S. (2000) · Cited by 3,479
Self-citationLearning and development in neural networks: the importance of starting small
Elman, J. L. (1993) · Cognition · doi:10.1016/0010-0277(93)90058-4 · Cited by 1,988 · cited 8×
“The idea of training a learning machine with a curriculum can be traced back at least to Elman (1993).”
BackgroundMethodInfluentialRestricted Boltzmann machines for collaborative filtering
Salakhutdinov, R., Mnih, A., & Hinton, G. (2007) · Proceedings of the 24th international conference on Machine learning · doi:10.1145/1273496.1273596 · Cited by 1,885
Research works printed but not linked: 3
- 10.5555/89603 · doi:10.5555/89603
- 10.5555/866740 · doi:10.5555/866740
- 10.5555/525571 · doi:10.5555/525571
Cited by
Items 2,901–3,000 of 5,240
Progressive Growing of Patch Size: Resource-Efficient Curriculum Learning for Dense Prediction Tasks
Fischer, S. M., Felsner, L., Osuala, R. O., Kiechle, J., Lang, D. M., Peeken, J. C., & Schnabel, J. A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72114-4_49 · Cited by 10
Advancing Brain Imaging Analysis Step-by-Step via Progressive Self-paced Learning
Yang, Y., Chen, H., Hu, J., Guo, X., & Ma, T. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72120-5_6 · Cited by 5
Wang, W., Cai, D., Liu, Z., et al. (2024) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc60668.2024.10694656 · Cited by 4
Chevalier, M., Trochut, S., Guizzetti, R., Urard, P., Labrak, L., Samuel, J., Cellier, R., & Abouchi, N. (2024) · International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design · doi:10.1109/smacd61181.2024.10745428 · Cited by 5
Deep Neural Network Compression for Lightweight and Accurate Fish Classification
Salie, D., Brown, D., & Chieza, K. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-78255-8_18 · Cited by 4
Gomaa, A., Mahdy, B., Kleer, N., & Krüger, A. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10802574 · Cited by 9
“1) We propose a novel approach for customizing the surgical behaviour of an imitation learning agent based on visual input and recorded demonstrations, tailoring it to individual surgeon techniques through Transfer Learning [14] and Curriculum Learning techniques [15].”
InfluentialSpinelli, F., Egli, P., Nubert, J., Nan, F., Bleumer, T., Goegler, P., Brockes, S., Hofmann, F., & Hutter, M. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10802199 · Cited by 10
“Both these terms are scaled and reshaped with curriculum learning [36], a technique which has become extremely popular in RL [37] because it allows agents to acquire complex skills by being tasked with environments of increasing difficulty.”
Unsupervised Gait Recognition With Selective Fusion
Ren, X., Yang, S., Hou, S., Cao, C., Liu, X., & Huang, Y. (2025) · IEEE Transactions on Biometrics Behavior and Identity Science · doi:10.1109/tbiom.2025.3544805 · Cited by 7
“CPC [41] uses curriculum learning [42] strategy to incorporate easy and hard samples and gradually relax the clustering criterion.”
Incorporating difference information into curriculum learning for multitemporal image classification
Zhao, Y., Li, H., Gong, M., et al. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127070 · Cited by 4
Dimension-Variable Mapless Navigation With Deep Reinforcement Learning
Zhāng, W., Zhang, Y., Liu, N., Ren, K., & Peng, G. (2025) · IEEE/ASME Transactions on Mechatronics · doi:10.1109/tmech.2025.3541797 · Cited by 6
“To address this problem, we follow the curriculum learning paradigm [31]: Start from simple navigation tasks and gradually increase the task difficulty.”
Pseudo-label Based Domain Adaptation for Zero-Shot Text Steganalysis
Luo, Y., Yang, Z., Zhang, R., & Liu, J. (2025) · Mechanisms and machine science · doi:10.1007/978-3-031-82907-9_10 · Cited by 5
SPCNet: Deep Self-Paced Curriculum Network Incorporated With Inductive Bias
Zhao, Y., Gong, M., Zhang, M., Qin, A. K., Jiang, F., & Li, J. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3544724 · Cited by 9
“CL, as proposed by Bengio et al. [16], is inspired by the learning principle of human beings that gradually incorporates knowledge from easy to complex.”
InfluentialKnowledge-Enhanced Deep Reinforcement Learning for Multi-Agent Game
Zeng, W., Yan, X., Mo, F., et al. (2025) · Electronics · doi:10.3390/electronics14071347 · Cited by 4
Characterization of Human Balance through a Reinforcement Learning-based Muscle Controller
Akbaş, K., Mummolo, C., & Zhou, X. A. (2025) · PLoS ONE · doi:10.1371/journal.pone.0320211 · Cited by 5
“CL [40] has been used in literature to learn complex human and robot movement skills [22, 23].”
Suhas, M., Abisset‐Chavanne, E., & Rey, P.-A. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127513 · Cited by 6
Co-Pseudo Labeling and Active Selection for Fundus Single-Positive Multi-Label Learning
Hu, T., Zhang, W., Guo, J., & Li, H. (2025) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2025.3565000 · Cited by 5
“For more precise supervision, Cascante-Bonilla et al. [34] boost the performance of pseudo-labeling by introducing curriculum learning [35] with a self-pacing strategy, instead of using vanilla handpicked thresholding.”
Acosta-Bermejo, R., Terrazas-Chavez, J. A., & Anaya, E. A. (2025) · Applied Sciences · doi:10.3390/app15179252 · Cited by 4
A curriculum learning approach to training antibody language models
Burbach, S. M., & Briney, B. S. (2025) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1013473 · Cited by 9
Enhancing IoT Network Security through Adaptive Curriculum Learning and XAI
Narkedimilli, S., Makam, S., Sriram, A. V., Mallellu, S. P., Sathvik, M., & Prasad, R. R. V. (2025) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf66845.2025.11317635 · Cited by 13
“Curriculum Learning [6] [7] is based on human education, a machine-learning approach in which models are trained in stages, first with simpler tasks and then with higher-level tasks.”
Text-to-video: a semantic search engine for internet videos
Jiang, L., Yu, S.-I., Meng, D., Mitamura, T., & Hauptmann, A. G. (2015) · International Journal of Multimedia Information Retrieval · doi:10.1007/s13735-015-0093-0 · Cited by 5
A HMM-based pre-training approach for sequential data.
Pasa, L., Testolin, A., & Sperduti, A. (2014) · Cited by 5
Multi-staged deep learning with created coarse and appended fine categories
Hagawa, R., Ishii, Y., & Tsukizawa, S. (2015) · IAPR Asian Conference on Pattern Recognition (ACPR) · doi:10.1109/acpr.2015.7486461 · Cited by 4
ViSeR: Visual Self-Regularization
Izadinia, H., & Garrigues, P. J. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00479 · Cited by 3
“Contrary to these approaches, we propose a form of curriculum learning [4] where we first train a model on a small set of clean data, and then augment the training set by mining instances from a large set of unlabeled images.”
MethodMulti-Scenario Deep Learning for Multi-Speaker Source Separation
Zegers, J., hamme, H. V., & Leuven, K. (2018) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2018.8461788 · Cited by 4
“The networks were trained using curriculum learning [17], i.”
MethodAutomatic Model Construction for Activity Recognition Using Wearable Devices
Bao, Y., & Chen, W. (2018) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomw.2018.8480411 · Cited by 4
“In this paper, we introduce an automatic model construction method with self-paced learning regime [4] for the wearable devices to automatically construct the activity recognition ability.”
BackgroundMethodAutomating Vehicles by Deep Reinforcement Learning Using Task Separation with Hill Climbing
Plessen, M. G. (2019) · Lecture notes in networks and systems · doi:10.1007/978-3-030-12385-7_16 · Cited by 12
“2) The problematic of curriculum learning: In [33], curriculum learning (CL) is discussed as a method to speed up learning by providing the learning agent first with simpler examples before gradually increasing complexity.”
BackgroundMethodInfluentialThe Balanced Loss Curriculum Learning
Qin, W., Hu, Z., Liu, X., Fu, W., He, J., & Hong, R. (2020) · IEEE Access · doi:10.1109/access.2020.2970726 · Cited by 4
“And then we exploit the predictions of the teacher model as boundary information to guide the 25994 VOLUME 8, 2020 Algorithm 1Detail Training Steps of Balanced Loss Curriculum Learning Framework (BLCL) Input: Training dataset (X ,Y); learning rate α; the hyperparameter r ; Output: The teacher network θs; The student n…”
BackgroundInfluentialFSPMTL: Flexible Self-Paced Multi-Task Learning
Sun, L., & Zhou, Y. (2020) · IEEE Access · doi:10.1109/access.2020.3009988 · Cited by 4
“Curriculum Learning (CL) mimics the cognitive process of humans and favors a learning algorithm to follow the logical learning sequence from simple examples to more difficult ones [22].”
BackgroundDeep curriculum learning optimization
Ghebrechristos, H., & Alaghband, G. (2020) · SN Computer Science · doi:10.1007/s42979-020-00251-7 · Cited by 6
“Most curriculum learning techniques get their inspiration from human learning where training is highly organized, based on education system and a curriculum which usually enables learning concepts in gradually increasing levels of difficulty while considering previously learned concepts [1].”
BackgroundResultAdaptive Transmission for Edge Learning via Training Loss Estimation
Huang, X., & Zhou, S. (2020) · International Conference on Communications · doi:10.1109/icc40277.2020.9149251 · Cited by 5
“Yet the strict definition of easy dataset and hardness is not given in [15].”
BackgroundAssessing Lesion Segmentation Bias of Neural Networks on Motion Corrupted Brain MRI
Mathai, T. S., Wang, Y., & Cross, N. M. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-72084-1_14 · Cited by 4
Generalization Bounds for Randomized Learning with Application to Stochastic Gradient Descent
London, B. (2017) · Cited by 4
Data Adaptive Image Enhancement and Classification for Synthetic Aperture Sonar
Gerg, I. D., Williams, D. P., & Monga, V. (2020) · IEEE International Geoscience and Remote Sensing Symposium (IGARSS) · doi:10.1109/igarss39084.2020.9324047 · Cited by 4
“In some sense, this can be seen as a form of curriculum learning [14] where samples are ignored as they become easier to classify resulting in focus on the more difficult and distorted samples as the training progresses.”
BackgroundOLALA: Object-Level Active Learning for Efficient Document Layout Annotation
Shen, Z., Li, W., Zhao, J. G., Yu, Y., & Dell, M. (2022) · Workshop on Natural Language Processing and Computational Social Science (NLP+CSS) · doi:10.18653/v1/2022.nlpcss-1.19 · Cited by 12
“According to Curriculum Learning (Bengio et al., 2009), we set high initial values of r to rely more on human labeling and ease model training in the beginning.”
MethodG2Basy: A framework to improve the RNN language model and ease overfitting problem
Yuwen, L., Chen, S., & Yuan, X. (2021) · PLoS ONE · doi:10.1371/journal.pone.0249820 · Cited by 5
Attention-Enhanced Gradual Machine Learning for Entity Resolution
Zhong, P., Li, Z., Chen, Q., et al. (2021) · IEEE Intelligent Systems · doi:10.1109/mis.2021.3077265 · Cited by 4
“Curriculum learning (CL) [8] and self-paced learning (SPL) [9] are to some extent similar to gradual machine learning in that they generally start with learning easier aspects of a task, and then gradually takes more complex examples into consideration.”
BackgroundSound Event Detection Based on Curriculum Learning Considering Learning Difficulty of Events
Tonami, N., Imoto, K., Okamoto, Y., et al. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414184 · Cited by 4
“To utilize the difference in the difficulty of training between the sound events, we employ curriculum learning [17].”
MethodGeneralizing Curricula for Reinforcement Learning
Narvekar, S., & Stone, P. (2020) · Cited by 9
Learning to Balance Local Losses via Meta-Learning
Yoa, S., Jeon, M., Oh, Y., & Kim, H. J. (2021) · IEEE Access · doi:10.1109/access.2021.3113934 · Cited by 4
“Curriculum learning [10], [11] and self-paced learning [12], [13] maximize learning efficiency by learning samples in a meaningful order such as an easy-to-hard order.”
BackgroundNoise tolerant ternary weight deep neural networks for analog in-memory inference
Doevenspeck, J., Vrancx, P., Laubeuf, N., Mallik, A., Debacker, P., Verkest, D., Lauwereins, R., & Dehaene, W. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9533684 · Cited by 10
Exploiting Curriculum Learning in Unsupervised Neural Machine Translation
Lu, J., & Zhang, J. (2021) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2021.findings-emnlp.79 · Cited by 12
“Curriculum learning (Bengio et al., 2009) is motivated by the learning strategy of biological organisms which orders the training samples in an easy-to-hard manner (Elman, 1993).”
BackgroundGu, Y., Hu, Q., Wang, X., Zhou, Z., & Lu, S. (2021) · International Congress on Image and Signal Processing (CISP) · doi:10.1109/cisp-bmei53629.2021.9624434 · Cited by 6
MIRecipe: A Recipe Dataset for Stage-Aware Recognition of Changes in Appearance of Ingredients
Zhang, Y., Yamakata, Y., & Tajima, K. (2021) · ACM International Conference on Multimedia in Asia (MMAsia) · doi:10.1145/3469877.3490596 · Cited by 4
“The original curriculum learning method [1] proposes to start with a sub-dataset consisting of simpler data, and then switch to a harder target dataset.”
MethodA Simple Approach to Balance Task Loss in Multi-Task Learning
Liang, S., Deng, C. H., & Zhang, Y. (2021) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata52589.2021.9671640 · Cited by 10
“The idea of curriculum learning [15] and its variant, i . e ., self-paced learning [18], have been adopted in multi-task learning [14], [16], [17] by firstly learning from easy tasks and then from harder ones.”
BackgroundEmotion-aware Human–Robot Interaction and Social Robots
Tian, L., Oviatt, S. L., Muszyński, M., Chamberlain, B. C., Healey, J., & Sano, A. (2022) · ACM eBooks · doi:10.1145/3502398.3502407 · Cited by 6
Learning camera invariant deep features for semi-supervised person re-identification
Zhu, H., Huang, L., Wei, Z., Zhang, W., & Cai, H. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-12581-0 · Cited by 4
“Curriculum Learning (CL) was proposed long age in [2, 19] take CL as a regularization term and propose self-paced Learning (SPL).”
BackgroundSelf-Communicating Deep Reinforcement Learning Agents Develop External Number Representations
Sabathiel, S., Solstad, T., Testolin, A., & Petruzzellis, F. (2022) · Proceedings of the Northern Lights Deep Learning Workshop · doi:10.7557/18.6291 · Cited by 5
“We designed a curriculum learning approach [5], where the agent learns simpler concepts (small numbers) first and progressively learns harder concepts (larger numbers).”
MethodDynamically throttleable neural networks
Liu, H., Parajuli, S., Hostetler, J., Chai, S. M., & Bhanu, B. (2022) · Machine Vision and Applications · doi:10.1007/s00138-022-01311-z · Cited by 10
“1) is adopted via curriculum learning [68].”
Specification Aware Multi-Agent Reinforcement Learning
Ritz, F., Phan, T., Müller, R., Gabor, T., Sedlmeier, A., Zeller, M., Wieghardt, J., Schmid, R., Sauer, H., Klein, C., et al. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-10161-8_1 · Cited by 6
Fine-grained label learning in object detection with weak supervision of captions
Wang, X., Du, Y., Verberne, S., & Verbeek, F. J. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-13592-7 · Cited by 3
“Curriculum learning [3] aims to formalize the learning process of humans and animals and organize it in a meaningful order to train models.”
BackgroundAssessing evolutionary terrain generation methods for curriculum reinforcement learning
Howard, D., Munn, H., Dolcetti, D., Kannemeyer, J., & Robinson, N. (2022) · Genetic and Evolutionary Computation Conference (GECCO) · doi:10.1145/3512290.3528870 · Cited by 9
“Curriculum Learning (CL) [5, 24, 34] is a powerful reinforcement learning technique that engenders powerful behaviours by gradually increasing the difficulty of the task to be solved.”
BackgroundMethodRen, L., Wang, C., Yang, Y. F., & Cao, Z. (2021) · IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio54168.2021.9739440 · Cited by 6
“In order to solve the problem of reward sparseness, different from the curriculum learning [16][17] adopted in [13], the proposed method borrows the idea of the hierarchical reinforcement learning (HEL) [18][19].”
MethodCharacterizing Continual Learning Scenarios for Tumor Classification in Histopathology Images
Kaustaban, V., Ba, Q., Bhattacharya, I., Sobh, N., Mukherjee, S., Martin, J. C., Miri, M. S., Guetter, C., & Chaturvedi, A. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-16961-8_18 · Cited by 8
“The results (Table S2) were consistent with the hypothesis of curriculum learning that knowledge is better captured if harder tasks follow easier tasks [6].”
ResultSelf-paced annotations of crowd workers
Kang, X., Yu, G., Domeniconi, C., Wang, J., Guo, W., Ren, Y., Zhang, X., & Cui, L. (2022) · Knowledge and Information Systems · doi:10.1007/s10115-022-01759-5 · Cited by 3
“Inspired by the intrinsic learning capability of humans, self-paced learning is proposed recently that gradually proceeds from easy to more complex samplesintrainingprocessbasedontheideaofcurriculumlearning[1].”
BackgroundIndoor Path Planning for Multiple Unmanned Aerial Vehicles via Curriculum Learning
Park, J. M., & Park, K. (2022) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc55196.2022.9952572 · Cited by 5
“Additionally, to improve learning performance, curriculum learning [51] was conducted in two stages.”
BackgroundA Deep Curriculum Learner in an Active Learning Cycle for Polsar Image Classification
Mousavi, S. H., & Azimi, S. M. (2022) · IEEE International Geoscience and Remote Sensing Symposium (IGARSS) · doi:10.1109/igarss46834.2022.9884910 · Cited by 4
“This problem could be solved by curriculum learning (CL), which imitates the meaningful learning order in human[5].”
BackgroundAhn, W. J., Kim, D.-W., Kang, T., Pae, D.-S., & Lim, M. T. (2023) · Applied Sciences · doi:10.3390/app13020781 · Cited by 5
Merlo, A., Böckenhoff, D., Schilling, J., Lazerson, S., & Pedersen, T. S. (2023) · Nuclear Fusion · doi:10.1088/1741-4326/acc852 · Cited by 8
“During the physics regularization stage, a curriculum learning approach is used [45].”
MethodNeural clustering based on implicit maximum likelihood
Vardakas, G., & Likas, A. C. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08524-x · Cited by 6
“The intuition is that this slow-paced training procedure ([58, 59]) strives to learn and cluster the ‘‘easier’’ data points first, like those that are close to the cluster centers and then tries to learn and cluster ‘‘more difficult’’ data points away from the cluster centers.”
BackgroundSahoo, A. K., & Chakraverty, S. (2023) · East Asian Journal on Applied Mathematics · doi:10.4208/eajam.2022-115.300922 · Cited by 4
A Heuristics-Based Reinforcement Learning Method to Control Bipedal Robots
Qin, D., Zhang, G., Zhu, Z., Chen, T., Zhu, W., Rong, X., Xie, A., & Li, Y. (2023) · International Journal of Humanoid Robotics · doi:10.1142/s0219843623500135 · Cited by 6
Samples with Low Loss Curvature Improve Data Efficiency
Garg, I., & Roy, K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01943 · Cited by 18
DMH-CL: Dynamic Model Hardness Based Curriculum Learning for Complex Pose Estimation
Dai, Y., Chen, B., Gao, L., Song, J., & Shen, H. T. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3307935 · Cited by 5
“Besides, existing research [16], [17], [18], [19] suggests that organizing training examples in a meaningful order, that is, learning from simple examples and gradually moving on to difficult ones, would benefit the learning process not only for humans but also for machines.”
BackgroundMethodDeep Stochastic Logic Gate Networks
Kim, Y. (2023) · IEEE Access · doi:10.1109/access.2023.3328622 · Cited by 17
“Sampling techniques play a vital role in various machine learning applications, enabling the generation of representative samples from probability distributions [8], [9].”
BackgroundCurriculum Guided Domain Adaptation in the Dark
Jahan, C. S., & Savakis, A. E. (2023) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2023.3330829 · Cited by 4
“Bengio et al. [34] introduced curriculum learning as a method of training a model with increasingly complex data samples, to mimic the human learning process.”
BackgroundMethodInfluentialDynamic Pursuit-Evasion Scenarios With a Varying Number of Pursuers Using Deep Sets
Logiewa, R., Hoffmann, F., Govaers, F., & Koch, W. (2023) · IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI) · doi:10.1109/sdf-mfi59545.2023.10361514 · Cited by 4
“To improve learning speed and success percentage, curriculum learning [14] was used, where the learning process starts with easy tasks and on each success gets progressively more challenging.”
MethodUnilaterally Aggregated Contrastive Learning with Hierarchical Augmentation for Anomaly Detection
Wang, G., Wang, Y., Qin, J., Zhang, D., Bao, X., & Huang, D. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00634 · Cited by 12
“…learn low-level features while deep layers emphasize more on high-level task-related semantic features, motivated by curriculum learning (CL) [2], we set stronger augmentation strengths for deeper layers and vice versa, aiming at capturing the distinct representations of in-liers from…”
BackgroundMethodAdaptive Curriculum Learning With Successor Features for Imbalanced Compositional Reward Functions
Szőke, L., Shperberg, S., Holtz, J., & Allievi, A. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3387134 · Cited by 5
“In Curriculum Learning (CL) [24] [25], the objective is to craft curricula tailored to a specific target task or distribution of tasks for the agent’s training.”
BackgroundImproving Data Augmentation for Robust Visual Question Answering with Effective Curriculum Learning
Zheng, Y., Wang, Z., & Chen, L. (2024) · International Conference on Multimedia Retrieval · doi:10.1145/3652583.3657607 · Cited by 8
“The easy-to-hard spirit of CL [7] is inspired by the learning process of us humans.”
BackgroundAutomatic Detection of Road Subsurface Distress via Curriculum Learning: Learn Like an Expert
Yue, G., Liu, C., Li, Y., Du, Y., & Gao, Q. (2024) · Transportation Research Record Journal of the Transportation Research Board · doi:10.1177/03611981241248164 · Cited by 6
Enhancing deep reinforcement learning for scale flexibility in real-time strategy games
Lemos, M. L. H. D., Vieira, R. E. S., Tavares, A. R., Marcolino, L. S., & Chaimowicz, L. (2024) · Entertainment Computing · doi:10.1016/j.entcom.2024.100843 · Cited by 5
Peng, Z., Zhou, X. W., Zheng, L., Wang, Y., & Ma, J. (2024) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10801582 · Cited by 14
“One promising way to alleviate the above problem is curriculum learning [17]– [19].”
Flexible top-view human pose estimation for detection system via CNN
Go, R., & Aoki, Y. (2016) · IEEE Global Conference on Consumer Electronics (GCCE) · doi:10.1109/gcce.2016.7800406 · Cited by 4
“In this paper, we cope with two problems in top-view human pose estimation by introducing CNN learning strategy that we call Bounding-box Curriculum Learning (BCL) inspired by Curriculum Learning [2] and performing post processing that we call Recurrent Pose Estimation (RPE).”
MethodInfluentialTask Analysis for Teaching Cumulative Learners
Bieger, J., & Þórisson, K. R. (2018) · Lecture notes in computer science · doi:10.1007/978-3-319-97676-1_3 · Cited by 5
“This concept forms the basis of teaching approaches like shaping [20,3], scaffolding [27], and part-task training [23], as well as for many concepts of intrinsic motivation or “curiosity” [21].”
BackgroundFixing Mini-batch Sequences with Hierarchical Robust Partitioning
Wang, S., Bai, W., Lavania, C., & Bilmes, J. (2019) · International Conference on Artificial Intelligence and Statistics · Cited by 4
Wong, P.-K., Leung, K., & Wong, M. (2019) · Expert Systems with Applications · doi:10.1016/j.eswa.2019.06.012 · Cited by 5
Parameter Continuation Methods for the Optimization of Deep Neural Networks
Pathak, H. N., & Paffenroth, R. C. (2019) · International Conference on Machine Learning and Applications · doi:10.1109/icmla.2019.00268 · Cited by 9
“The smoothed objective function is minimized first, and then gradually smoothing is reduced and simultaneously minimized as training proceeds to obtain the final solution [4].”
InfluentialResearch on Task Discovery for Transfer Learning in Deep Neural Networks
Akdemir, A. (2020) · Annual Meeting of the Association for Computational Linguistics: Student Research Workshop (ACL SRW) · doi:10.18653/v1/2020.acl-srw.6 · Cited by 5
“In this context, we define difficulty as the amount of contextual information necessary to select the most probable word, whereas Bengio et al. (2009) defined difficulty as the inverse of the frequency of each masked token regardless of their contexts.”
BackgroundMethodInfluentialSteel surface defects classification method based on self-paced learning
Liang, D., Chen, D., Liu, S., Jia, X., & Zhao, C. (2020) · Chinese Control Conference (CCC) · doi:10.23919/ccc50068.2020.9189598 · Cited by 4
“SPL draws on the thinking of the process of human learning in the Curriculum learning (CL) [12].”
BackgroundAutomated optimal control in energy systems: the reinforcement learning approach
Zhang, X., & Jiang, H. (2020) · Elsevier eBooks · doi:10.1016/b978-0-12-820168-8.00015-8 · Cited by 5
Adaptively Leverage Unlabeled Tracklets Based on Part Attention Model for Few-Example Re-ID
Han, J., Li, Y., & Wang, S. (2020) · IEEE Signal Processing Letters · doi:10.1109/lsp.2020.3037473 · Cited by 4
Wachowiak, L., Lång, C., Heinisch, B., & Gromann, D. (2020) · Cited by 4
Partial Label Learning via Self-Paced Curriculum Strategy
Lyu, G., Feng, S., Jin, Y., et al. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-67661-2_29 · Cited by 4
“Self-paced Curriculum Strategy can be regard as an integration of curriculum strategy and self-paced strategy, which are separately proposed by [2] and [17].”
BackgroundSimulating Early Word Learning in Situated Connectionist Agents.
Hill, F., Clark, S. R. L., Blunsom, P., & Hermann, K. M. (2020) · eScholarship (California Digital Library) · Cited by 8
Getting a Head Start on Program Synthesis with Genetic Programming
Wick, J., Hemberg, E., & O’Reilly, U.-M. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-72812-0_17 · Cited by 4
“This is similar to curriculum learning [4] in ML which starts with a small task and then introduces of a multi-stage curriculum.”
LODENet: A Holistic Approach to Offline Handwritten Chinese and Japanese Text Line Recognition
Hoang, H.-T., Peng, C.-J., Tran, H. V., et al. (2021) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr48806.2021.9412161 · Cited by 4
“For efficient training, we derived a training procedure based on Curriculum Learning [32].”
MethodCurriculum Learning by Optimizing Learning Dynamics
Zhou, T., Wang, S., & Bilmes, J. (2021) · International Conference on Artificial Intelligence and Statistics · Cited by 37
Unveiling the pole structure of S-matrix using deep learning
Sombillo, D. L. B., Ikeda, Y., Sato, T., et al. (2022) · Suplemento de la Revista Mexicana de Física · doi:10.31349/suplrevmexfis.3.0308067 · Cited by 4
“For most classification problems, additional procedure is expected since one has to identify which part of the dataset is less complicated than the other [37, 38].”
BackgroundDistilling Reinforcement Learning Tricks for Video Games
Kanervisto, A., Scheller, C., Schraner, Y., & Hautamäki, V. (2021) · IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9618997 · Cited by 7
“Another common solution to sparse reward environments is to use curriculum learning to initially start with an easy problem and gradually increase the difficulty [19].”
Discretisation and continuity: The emergence of symbols in communication
Lieck, R., & Rohrmeier, M. (2021) · Cognition · doi:10.1016/j.cognition.2021.104787 · Cited by 5
You should know more: Learning external knowledge for visual dialog
Zhao, L., Zhang, H., Li, X., Yang, S., & Song, Y. (2022) · Neurocomputing · doi:10.1016/j.neucom.2021.10.121 · Cited by 6
Generalising from conventional pipelines using deep learning in high-throughput screening workflows
Cruz, B. G. S., Slter, J., Gomez‐Giro, G., Saraiva, C., Sabaté‐Soler, S., Modamio, J., Barmpa, K., Schwamborn, J. C., Hertel, F., Jarazo, J., et al. (2022) · Scientific Reports · doi:10.1038/s41598-022-15623-7 · Cited by 4
Deep learning strategy for salt model building
Klatt, M., Faria, E. L., Muller, A. P. O., Coelho, J. M., Gonzalez, J. L., de Albuquerque, M. P., de Albuquerque, M. P., Bom, C. R., & Correia, M. D. (2022) · Geophysics · doi:10.1190/geo2021-0362.1 · Cited by 4
Self-paced uncertainty estimation for one-shot person re-identification
Zhang, Y., Ma, B., Liu, L., Xin, Y., Li, M., & Diao, Y. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04245-1 · Cited by 5
Image-to-Height Domain Translation for Synthetic Aperture Sonar
Stewart, D., Kreulach, A., Johnson, S. F., & Zare, A. (2021) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3236473 · Cited by 7
“Future work should investigate curriculum learning strategies to obtain possible improvements of the approach [43].”
BackgroundCurriculum based Reinforcement Learning for traffic simulations
Makri, S., & Charalambous, P. (2023) · Computers & Graphics · doi:10.1016/j.cag.2023.04.009 · Cited by 4
Giving advice to agents with hidden goals
Rosman, B., & Ramamoorthy, S. (2014) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra.2014.6907118 · Cited by 5
“This approach is typically referred to as reward shaping or curriculum learning [14], [15], [16], [17].”
BackgroundReducing the Effects of Detrimental Instances
Smith, M. R., & Martinez, T. R. (2014) · International Conference on Machine Learning and Applications · doi:10.1109/icmla.2014.34 · Cited by 4
“This procedure has developed into curriculum learning and has had success in deep learning [2].”
MethodAn Object is Worth Six Thousand Pictures: The Egocentric, Manual, Multi-image (EMMI) Dataset
Wang, X., Eliott, F. M., Ainooson, J., Palmer, J. H., & Kunda, M. (2017) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2017.279 · Cited by 9
“Perhaps the largest bodies of relevant work are in active learning [24] and curriculum learning [5].”
BackgroundVietnamese Service Robot Based on Artificial Intelligence
Phong, N. T. T., Nam, L. H. T., & Thinh, N. T. (2020) · International Journal of Mechanical Engineering and Robotics Research · doi:10.18178/ijmerr.9.5.701-708 · Cited by 4
Metadata
- Authors
- Bengio, Y., Louradour, J., Collobert, R., & Weston, J.
- Affiliations
- Université de Montréal (CA)
- NEC Laboratories America, Princeton, NJ
- ORCID
- Bengio, Yoshua
- Venue
- Proceedings of the 26th Annual International Conference on Machine Learning
- Publisher
- ACM
- Published
- 2009-06-14
- Pages
- 41-48
- OpenAlex
- W2296073425
- Semantic Scholar
- 8de174ab5419b9d3127695405efd079808e956e8
- Cited by
- 5,366
- References
- 40
- Citations by year
- 2026: 526 · 2025: 636 · 2024: 551 · 2023: 591 · 2022: 536 · 2021: 711 · 2020: 565 · 2019: 442 · 2018: 293 · 2017: 198 · 2016: 103 · 2015: 63 · 2014: 41 · 2013: 40 · 2012: 33
- FWCI
- 16.17
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- International Conference on Machine Learning (publishedVersion)
- Queen Mary Research Online (Queen Mary University of London) (submittedVersion)
- Funders
- Canadian Institute for Advanced Research
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Stochastic Gradient Optimization Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Neural Networks and Applications
- Keywords
- curriculum learning, training order, non-convex optimization, deep neural network, stochastic neural networks, deterministic neural networks, generalization, convergence rate, numerical continuation, global optimization
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.