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 2,601–2,700 of 5,240
Rethinking Curriculum Learning with Incremental Labels and Adaptive CompensationPreprint
Ganesh, M. R., & Corso, J. J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2001.04529 · Cited by 6
“There are two alternative lines of work which offer strategies to guide deep networks to better solutions than batch learning: Curriculum Learning [3, 12, 15] and Label Smoothing [9, 38].”
BackgroundMethodInfluentialLAMP: Large Deep Nets with Automated Model Parallelism for Image SegmentationPreprint
Zhu, W., Zhao, C., Li, W., et al. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.12575 · Cited by 6
“learning theory such as curriculum learning [1,12], a model can be well trained by firstly being fit easy samples/tasks and later being fit hard samples/tasks.”
BackgroundSymbolic Relational Deep Reinforcement Learning based on Graph Neural Networks.Preprint
Janisch, J., Pevný, T., & Lisý, V. (2020) · arXiv (Cornell University) · Cited by 8
Ehrlich, D. B., Stone, J. T., Brandfonbrener, D., et al. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.09.30.321752 · Cited by 6
Anti-Distillation: Improving reproducibility of deep networksPreprint
Shamir, G. I., & Coviello, L. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.09923 · Cited by 19
“While in some way, this mimics real life, where individuals make conclusions based on what they learned and what they know, or dependent on the order in which they learn different topics (see, e.g., Achille et al. (2017); Bengio et al. (2009)), this is definitely not a desired behavior.”
BackgroundDeep Model-Based Reinforcement Learning for High-Dimensional Problems, a SurveyPreprint
Plaat, A., Kosters, W. A., & Preuß, M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.05598 · Cited by 58
“Short-term future frames are predicted and fine-tuned to predict longerterm future frames after the previous phase converges, using a curriculum that stabilizes training [Bengio et al., 2009].”
BackgroundMethodInfluentialMontella, S., Fabre, B., Urvoy, T., Heinecke, J., & Rojas-Barahona, L. M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.00571 · Cited by 19
“In (Bengio et al., 2009), more complex samples are progressively added during training.”
InfluentialSOAR: Second-Order Adversarial RegularizationPreprint
Ma, A., Faghri, F., Papernot, N., & Farahmand, A. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2004.01832 · Cited by 17
Petitpierre, R. (2020) · arXiv (Cornell University) · doi:10.13140/rg.2.2.10973.64484 · Cited by 19
Distributed Heuristic Multi-Agent Path Finding with CommunicationPreprint
Ma, Z., Luo, Y., & Ma, H. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.11365 · Cited by 6
“Instead, we use curriculum learning method by gradually introducing more difficult tasks to agents [20].”
A Simultaneous Denoising and Dereverberation Framework with Target DecouplingPreprint
Li, A., Liu, W., Luo, X., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.12743 · Cited by 6
“Inspired by the curriculum learning concept [13], multistage SE begin to thrive in the recent several years [11, 14, 15, 16, 17].”
MethodInformation-Theoretic Measures of Dataset Difficulty.Preprint
Ethayarajh, K., Choi, Y., & Swayamdipta, S. (2021) · arXiv (Cornell University) · Cited by 9
Okamoto, W., Kera, H., & Kawamoto, K. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.10005 · Cited by 22
“Curriculum learning [15] is a learning process in which the difficulty of a task is gradually increased during training.”
BackgroundSeurin, P., & Shirvan, K. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4673021 · Cited by 6
OLALA: Object-Level Active Learning for Efficient Document Layout Annotation
Shen, Z., Zhao, J., Dell, M., Yu, Y., & Li, W. (2020) · arXiv · doi:10.48550/arxiv.2010.01762 · Cited by 6
Confidence-Aware Mean Teacher for semi-supervised metallographic image semantic segmentation
Cao, Y., Luo, B., Chen, Y., Xu, L., & Ding, C. (2025) · Computational Materials Science · doi:10.1016/j.commatsci.2024.113645 · Cited by 9
Curriculum learning with class-label composition for weakly supervised semantic segmentation
Hwang, D., Kim, H., Baek, D., Kim, H., Kye, I., & Choe, J. (2025) · Pattern Recognition Letters · doi:10.1016/j.patrec.2024.12.016 · Cited by 7
Transformer models for quantum gate set tomography
Yu, K. Y., Sarkar, A., Rimbach-Russ, M., et al. (2025) · Quantum Machine Intelligence · doi:10.1007/s42484-025-00237-9 · Cited by 5
“…the way humans learn, curriculum learning aims to achieve better performance and faster convergence by starting with simpler or more fundamental examples and progressively introducing more complex ones [27], we conveniently call it stage training here, starting from easier to more difficult stages.”
Causally-Aware Reinforcement Learning for Joint Communication and Sensing
Roy, A., Banerjee, S., Sadasivan, J., et al. (2025) · IEEE Transactions on Machine Learning in Communications and Networking · doi:10.1109/tmlcn.2025.3562557 · Cited by 5
Brain-Guided Self-Paced Curriculum Learning for Adaptive Human-Machine Interfaces
Choi, Y.-W., Shin, H.-B., & Lee, S.-W. (2025) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2025.3560832 · Cited by 9
“This neural mechanism inspired the introduction of curriculum learning in machine learning, which was first introduced by Bengio et al. [12], and led to the emergence of various curriculum-based algorithms [13], [14], [15], [16], [17].”
DP-PINN+: A Dual-Phase PINN learning with automated phase division
Yan, D., & He, L. (2025) · Journal of Computational Science · doi:10.1016/j.jocs.2025.102637 · Cited by 5
He, X., Pan, Y., Chen, Y., Li, Z., Xu, Z., Yang, C., & Wang, K. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3597474 · Cited by 5
“CL, first introduced by Bengio et al. [23], is inspired by educational methods, where learning begins with simpler tasks and gradually progresses to more complex ones.”
A DRL Framework for Autonomous Pursuit-Evasion: From Multi-Spacecraft to Multi-Drone Scenarios
Xu, Z., Shao, S., & Han, Z. (2025) · Drones · doi:10.3390/drones9090636 · Cited by 5
“The task difficulty is then gradually increased over the course of the training [33].”
Semi-Supervised Text-Based Person Search
Gao, D., Bai, Y., Cao, M., Dou, H., Ye, M., & Zhang, M. (2025) · IEEE Transactions on Image Processing · doi:10.1109/tip.2025.3607637 · Cited by 11
“To investigate the influence of applying di ff erent scheduling strategies, we experiment with another two popular training schedulers including Baby step [41], [53] and Root-2 [50].”
Sheshanarayana, R., & You, F. (2025) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.5c01508 · Cited by 5
Open-ended coordination for multi-agent systems using modular open policies
Rother, D., Pajarinen, J., Peters, J., & Weisswange, T. H. (2025) · Autonomous Agents and Multi-Agent Systems · doi:10.1007/s10458-025-09723-7 · Cited by 8
“Curriculum learning strategies have been employed to improve training efficiency by gradually increasing task difficulty [ 32 ].”
Policy Search through Genetic Programming and LLM-assisted Curriculum Learning
Jorgensen, S., Nadizar, G., Pietropolli, G., Manzoni, L., Medvet, E., O’Reilly, U.-M., & Hemberg, E. (2025) · ACM Transactions on Evolutionary Learning and Optimization · doi:10.1145/3772718 · Cited by 7
“It was first introduced by Elman [1993] and then formally defined by Bengio et al. [2009].”
MambaNUT: Nighttime UAV Tracking via Mamba-based Adaptive Curriculum Learning
Wu, Y., Yang, X., Wang, X., Ye, H., Zeng, D., & Li, S. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247071 · Cited by 16
“The concept of curriculum learning (CL), first proposed in [31], shows that the strategy of learning from easy to hard significantly enhances the generalization of deep models.”
Investigating Uncertainty Weighting for Multi-Task Learning: Insights and Analytical Alternative
Kirchdorfer, L., Sesterhenn, T., Bartelt, C., Stuckenschmidt, H., Schott, L., & Köhler, J. M. (2025) · International Journal of Computer Vision · doi:10.1007/s11263-025-02625-x · Cited by 18
Large Language Model-based power dispatch agent: Framework, application and challenges
Zhao, H. Q., Cheng, Y., Xiang, D., Zhou, X., Zhao, J., Cai, X., & Dong, Z. Y. (2026) · International Journal of Electrical Power & Energy Systems · doi:10.1016/j.ijepes.2026.111653 · Cited by 9
Nagano, M., & Nakamura, T. (2023) · Advanced Robotics · doi:10.1080/01691864.2023.2252048 · Cited by 5
“Therefore, we plan to introduce curriculum learning [64,65], which changes the complexity of the data used in the learning process.”
MethodMARR: A Multi-Agent Reinforcement Resetter for Redirected Walking
Lee, H. J., Jeon, S.-B., Cho, Y.-H., & Lee, I. (2024) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2024.3368043 · Cited by 9
Uduagbomen, J., Leeson, M. S., Liu, Z., Lakshminarayana, S., & Xu, T. (2024) · Applied Optics · doi:10.1364/ao.524426 · Cited by 5
Mouse2Vec: Learning Reusable Semantic Representations of Mouse Behaviour
Zhang, G., Hu, Z., Bâce, M., & Bulling, A. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642141 · Cited by 13
“We adopted a curriculum learning strategy [6], initially setting a low dropping ratio of 0.3 and gradually increasing it to 0.8.”
MethodInfluentialBeyond-Voice: Towards Continuous 3D Hand Pose Tracking on Commercial Home Assistant Devices
Li, Y., Reddy, R., Zhang, C., & Nandakumar, R. (2024) · ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1109/ipsn61024.2024.00017 · Cited by 10
“Therefore, we leverage curriculum learning (CL)[4] to hierarchically bootstrap the training by allowing the model to start with learning simple tasks before complex tasks.”
Crowd Perception Communication-Based Multi- Agent Path Finding With Imitation Learning
Xie, J., Zhang, Y., Yang, H., Ouyang, Q., Dong, F., Guo, X., Jin, S., & Shi, D. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3455948 · Cited by 10
“And scenarios are generated by curriculum learning [38] (details see Section V-A).”
BackgroundZhou, Z., Feng, Y., Qiu, A., Duan, G., & Zhou, M. (2024) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2024.3480014 · Cited by 7
“The original concept of CL was first proposed by Bengio et al. [44], which mimics the way humans learn.”
BackgroundPajić, Z., Janković, Z., & Selakov, A. (2024) · Energies · doi:10.3390/en17205183 · Cited by 5
Curriculum Prompting Foundation Models for Medical Image Segmentation
Zheng, X., Zhang, Y., Zhang, H., Liang, H., Bao, X., Jiang, Z., & Lao, Q. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72390-2_46 · Cited by 13
Enhancing LoRa Reception with Generative Models: Channel-Aware Denoising of LoRaPHY Signals
Sabharwal, K., Ramesh, S., Wang, J., Divakaran, D. M., & Chan, M. C. (2024) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/3666025.3699354 · Cited by 7
“Curriculum Learning: The training regimen adopts a curriculum learning approach [3].”
Active Generation for Image Classification
Huang, T., Liu, J., You, S., & Xu, C. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-73195-2_16 · Cited by 13
Lin, W.-C., Sridhar, K., & Busso, C. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3507562 · Cited by 7
Enhanced Graph Transformer: Multi-scale attention with Heterophilous Curriculum Augmentation
Zhuang, J., Li, J., Shi, C., Lin, X., & Fu, Y.-G. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.112874 · Cited by 6
Curriculum Learning for Self-Iterative Semi-Supervised Medical Image Segmentation
Zhang, J., Xu, D.-X., Chen, Y.-Y., Lou, Y., & Huang, Y. (2024) · IEEE International Conference on Bioinformatics and Biomedicine · doi:10.1109/bibm62325.2024.10822056 · Cited by 5
Auxiliary Reward Generation With Transition Distance Representation Learning
Li, S., Han, S., Zhao, Y., Yang, Y., Zhao, Q., & Liu, P. (2025) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2025.3554861 · Cited by 8
“Curriculum RL [7] tackles reward generation from a di ff erent view from reward shaping, which gradually learns policies in an automatically generated task sequence with increasing di ffi culty.”
A data-centric chip design agent framework for Verilog code generation
Chang, K., Zhu, W., Wang, K., He, X., Yang, N., Chen, Z., Jin, D., Li, C., Zhou, Y., Yan, H., et al. (2025) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3727980 · Cited by 13
“In a single fine-tuning epoch, the order of data presentation significantly impacts model accuracy [2], which is characterized to the curriculum learning [32].”
Spectral Adaptive Dropout: Frequency-Based Regularization for Improved Generalization
Huang, Z., Chen, M., & Zheng, S. (2025) · Information · doi:10.3390/info16060475 · Cited by 11
Generative attack in complex real-world scenarios
Peng, H., Cheng, G., & Sun, X. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.111893 · Cited by 4
Environment as Policy: Learning to Race in Unseen Tracks
Wang, H., Xing, J., Messikommer, N., & Scaramuzza, D. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127376 · Cited by 14
“However, training directly on a wide distribution of scenarios can significantly degrade the efficiency and effectiveness of the learning process [15].”
Battista, A., Padoa‐Schioppa, C., & Wang, X.-J. (2026) · Neuron · doi:10.1016/j.neuron.2025.12.010 · Cited by 8
Leveraging Noisy Online Databases For Use In Chord Recognition.
McVicar, M., Ni, Y., Bie, T. D., et al. (2011) · doi:10.5281/zenodo.1418310 · Cited by 5
Guarantees for Approximate Incremental SVMs
Usunier, N., Bordes, A., & Bottou, L. (2010) · Cited by 5
Zhu, Q., Xu, N., Huang, S.-J., Qian, J., & Zhang, D. (2019) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-019-00986-7 · Cited by 5
“Motivated by human learning, curriculum learning (CL) [20] mechanism has been introduced into machine learning, which gradually proceeds from easy to more complex samples in the training model.”
BackgroundBeyond Error Propagation: Language Branching Also Affects the Accuracy of Sequence Generation
Wu, L., Tan, X., Qin, T., Lai, J., & Liu, T. (2019) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2019.2933727 · Cited by 5
“probability) of a target language, which has been used as a kind of characterization for the hardness of learning [7], and the dependency statistics in parse trees.”
BackgroundCrowd Teaching with Imperfect Labels
Zhou, Y., Nelakurthi, A. R., Maciejewski, R., Fan, W., & He, J. (2020) · International World Wide Web Conference (WWW) · doi:10.1145/3366423.3380099 · Cited by 15
“However, in real-world teaching, a more effective strategy is following the principle of curriculum learning [2], which encourages the recommended teaching sequence to have tasks that range from easy to difficult.”
BackgroundPhonetic and Visual Priors for Decipherment of Informal Romanization
Ryskina, M., Gormley, M. R., & Berg-Kirkpatrick, T. (2020) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2020.acl-main.737 · Cited by 9
“and curriculum learning (Bengio et al., 2009) to achieve faster convergence, as described in §3.”
BackgroundMethodInfluentialFeature Aggregation Networks for Image Steganalysis
Jang, H., Oh, T., & Kim, K. (2020) · ACM Workshop on Information Hiding and Multimedia Security (IH&MMSec) · doi:10.1145/3369412.3395072 · Cited by 5
“The curriculum learning [1] was used to facilitate model learning.”
MethodDeep Likelihood Network for Image Restoration With Multiple Degradation Levels
Guo, Y., Lu, M., Zuo, W., Zhang, C., & Chen, Y. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3051767 · Cited by 7
“Besides multi-task learning, curriculum learning [11] and self-paced learning [12] can also be adopted to mitigate the issue by controlling the sampling strategies [9], since deep models are believed to learn much better if the training samples corresponding to tasks of different difficulties are…”
BackgroundBoosting Multilabel Semantic Segmentation for Somata and Vessels in Mouse Brain
Wu, X., Tao, Y., He, G., Liu, D., Meiling, F., Yang, S., Gong, H., Xiao, R., Chen, S., & Huang, J. (2021) · Frontiers in Neuroscience · doi:10.3389/fnins.2021.610122 · Cited by 4
“The issue could be largely leveraged by using a curriculum strategy (Bengio et al., 2009).”
BackgroundQuantifying Appropriateness of Summarization Data for Curriculum Learning
Kano, R., Takahashi, T., Nishino, T., Taniguchi, M., Taniguchi, T., & Ohkuma, T. (2021) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2021.eacl-main.119 · Cited by 10
“Curriculum learning is a method to change the order of training data to improve convergence speed and accuracy (Bengio et al., 2009).”
BackgroundMethodTask-Coupling Elastic Learning for Physical Sign-Based Medical Image Classification
Xu, Y., Wen, G., Yang, P., Fan, B., Hu, Y., Luo, M., & Wang, C. (2021) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2021.3106837 · Cited by 5
Solving the Shepherding Problem: Imitation Learning Can Acquire the Switching Algorithm
Go, C. K., Koganti, N., & Ikeda, K. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9533722 · Cited by 9
POPCORN: Progressive Pseudo-Labeling with Consistency Regularization and Neighboring
Kamraoui, R. A., Ta, V., Papadakis, N., Compaire, F., Manjón, J. V., & Coupé, P. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-87196-3_35 · Cited by 9
“Curriculum learning [3] showed that presenting data with an increasing difficulty can lead to a better learning process.”
BackgroundLearning to Hallucinate Examples from Extrinsic and Intrinsic Supervision
Gui, L., Bardes, A., Salakhutdinov, R., Hauptmann, A. G., Hebert, M., & Wang, Y.-X. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00858 · Cited by 6
“To address this issue, we propose a progressive guidance scheme inspired by curriculum learning [8], and explore two dual directions – (1) we start with a mentor and a student, both trained on a small number of real examples, and we gradually strengthen the mentor by re-training it with increasing…”
MethodInfluentialLung Cancer Prediction Using Curriculum Learning Based Deep Neural Networks
Zhou, J., Khushi, M., Moni, M. A., et al. (2021) · IEEE International Conference on Digital Health (ICDH) · doi:10.1109/icdh52753.2021.00013 · Cited by 5
CO-PILOT: COllaborative Planning and reInforcement Learning On sub-Task curriculum
Ao, S., Zhou, T., Long, G., Lu, Q., Zhu, L., & Jiang, J. B. (2021) · neural information processing systems · Cited by 20
A machine education approach to swarm decision-making in best-of-n problems
Hussein, A., Elsawah, S., Petraki, E., & Abbass, H. A. (2021) · Swarm Intelligence · doi:10.1007/s11721-021-00206-5 · Cited by 9
“As for sequencing the lessons for each skill, we use an iterative learning process where iteration complexity increases over time, as proposed in (Bengio et al. 2009).”
MethodA multichannel learning-based approach for sound source separation in reverberant environments
Chen, Y.-S., Lin, Z.-J., & Bai, M. R. (2021) · EURASIP Journal on Audio Speech and Music Processing · doi:10.1186/s13636-021-00227-2 · Cited by 6
Curriculum Generation and Sequencing for Deep Reinforcement Learning in StarCraft II
Hao, D., Sweetser, P., & Aitchison, M. (2022) · Australasian Computer Science Week Multiconference (ACSW) · doi:10.1145/3511616.3513093 · Cited by 6
“CL states that by exposing machine learning algorithms to tasks in meaningful orders, it can speed up convergence and can even lead to the discovery of better local minima [3].”
BackgroundYasutomi, A. Y., Hiroki, M. O. R. I., & Ogata, T. (2022) · IEEE/SICE International Symposium on System Integration (SII) · doi:10.1109/sii52469.2022.9708766 · Cited by 8
Supervised Attention in Sequence-to-Sequence Models for Speech Recognition
Yang, G.-P., & Tang, H. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746310 · Cited by 8
“In this paper, we further study this assumption with curriculum training [14], turning off the supervision midway through training.”
BackgroundCross-Target Stance Detection Via Refined Meta-Learning
Ji, H., Lin, Z., Fu, P., et al. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746302 · Cited by 5
TxVAD: Improved Video Action Detection by Transformers
Wu, Z., Ren, Z., Wu, Y., Wang, Z., & Hua, G. (2022) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3503161.3547992 · Cited by 5
“We locate one crucial cause: at initialization, the attention modules cast nearly uniform attention weights to all the cells in the feature maps.”
Cross-Modal Data Augmentation for Tasks of Different Modalities
Chen, D., Zhuang, Y., Shen, Z., Yang, C., Wang, G., Tang, S., & Yang, Y. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3228696 · Cited by 9
“This phenomenon can be explained by curriculum learning [54].”
BackgroundDeep Intelligence: What AI Should Learn from Nature’s Imagination
Minai, A. A. (2023) · Cognitive Computation · doi:10.1007/s12559-023-10124-9 · Cited by 8
“Unfortunately, there was very little follow-up on Elman’s suggestions until the recent emergence of curriculum learning [97], which has since seen significant growth [98, 99].”
Self-Supervised Learning from Untrimmed Videos via Hierarchical Consistency
Qing, Z., Zhang, S., Huang, Z., Xu, Y., Wang, X., Gao, C., Jin, R., & Sang, N. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3273415 · Cited by 10
“Curriculum learning [90] shows that models can learn much better when training examples are not randomly provided but organized in a meaningful order, i.”
BackgroundObscurity-Quantified Curriculum Learning for Machine Translation Evaluation
Zhang, C., Wong, D. F., Lei, E., Zhan, R., & Chao, L. S. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3282105 · Cited by 7
“Restrictions apply. area [31].”
Anime Character Identification and Tag Prediction by Multimodality Modeling: Dataset and Model
Fan, Y., Wu, J., Zhao, M., & Zhou, S. (2023) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn54540.2023.10191980 · Cited by 6
“When curriculum learning [19] was first proposed, it was an optimization strategy that centered on training from simple to difficult.”
BackgroundNon-Sequential Graph Script Induction via Multimedia Grounding
Zhou, Y., Li, S., Li, M., Lin, X., Chang, S.-F., Bansal, M., & Ji, H. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.303 · Cited by 11
Li, M., Zhou, T., Huang, Z., Yang, J. J., Yang, J., & Gong, C. (2023) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3635310 · Cited by 13
“Besides, Multi-part Curriculum learning Framework (MTCF) [61] proposes to distinguish ID data and OOD data in an ordered sequence from simple to difficult, and employ curriculum learning [1, 17] to achieve this purpose.”
BackgroundUnderstanding Translationese in Cross-Lingual Summarization
Wang, J., Meng, F., Liang, Y., Zhang, T., Xu, J., Li, Z., & Zhou, J. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.250 · Cited by 12
SelfRemaster: Self-Supervised Speech Restoration for Historical Audio Resources
Saeki, T., Takamichi, S., Nakamura, T., Tanji, N., & Saruwatari, H. (2023) · IEEE Access · doi:10.1109/access.2023.3345027 · Cited by 6
“Inspired by curriculum learning [55], we pretrain the analysis and channel modules in a supervised manner, before training our model with the methods described in § III-A to § III-D.”
MethodLearning to Learn: How to Continuously Teach Humans and Machines
Singh, P., Li, Y., Sikarwar, A., Lei, W., Gao, D., Talbot, M. B., Sun, Y., Shou, M. Z., Kreiman, G., & Zhang, M. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01075 · Cited by 15
“Previous works in curriculum learning can be categorized into predefined curriculum learning [8, 56, 12, 13] and automatic curriculum learning [62, 29, 16, 21].”
BackgroundResultLearning sample-aware threshold for semi-supervised learning
Wei, Q., Feng, L., Sun, H., Wang, R., He, R., & Yin, Y. (2024) · Machine Learning · doi:10.1007/s10994-023-06425-7 · Cited by 6
“Zhang et al. (2021) leveraged the idea of curriculum learning (Bengio et al. 2009) to take into account the learning status of each class and flexibly adjusted thresholds for different classes at each time step via a preset function.”
Pre-Grasp Approaching on Mobile Robots: A Pre-Active Layered Approach
Naik, L., Kalkan, S., & Krüger, N. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3358077 · Cited by 10
“Curriculum learning (CL) [19] is the form of TL wherein experience acquired by the agent over time is sorted in order to accelerate learning.”
BackgroundEasy-to-Hard Domain Adaptation With Human Interaction for Hyperspectral Image Classification
Zhang, C., Zhong, S., Wan, S., & Gong, C. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3358869 · Cited by 8
“In 2009, Bengio et al. [34] proposed CL, a learning approach that aims to imitate the human learning process, which can be viewed as a special kind of continuation method.”
Towards a better negative sampling strategy for dynamic graphs
Gao, K., Liu, C., Wu, J. X., Du, B., & Hu, W. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106175 · Cited by 7
Triplet Loss with Curriculum Learning for Audio-Visual Retrieval
Zeng, D., & Ikeda, K. (2023) · IEEE International Symposium on Multimedia (ISM) · doi:10.1109/ism59092.2023.00038 · Cited by 5
“In this paper, we present a novel approach rooted in curriculum learning [1] to bridge this gap and improve the performance of cross-modal retrieval models.”
MethodZhou, C., Song, D., Wu, Z., Hu, L., Xu, J., & Zhou, Y. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.112108 · Cited by 5
Fast FixMatch: Faster Semi-Supervised Learning with Curriculum Batch Size
Chen, J., Dun, C., & Kyrillidis, A. (2024) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit57864.2024.10619518 · Cited by 7
“Curriculum learning improves deep learning training performance by ordering training data using specific schemes into a “curriculum” [21], [22].”
BackgroundBoosting Pseudo-Labeling With Curriculum Self-Reflection for Attributed Graph Clustering
Zhu, P., Li, J., Wang, Y., Xiao, B., Zhang, J., Lin, W., & Hu, Q. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3416167 · Cited by 9
“Curriculum learning [47] assigns samples with different learning difficulties and trains all samples from easy to difficult.”
Research on Obstacle Avoidance Strategy of Automated Heavy Vehicle Platoon in High-Speed Scenarios
Cai, Y., Zhan, L., Sun, X., et al. (2024) · Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering · doi:10.1177/09544070241276062 · Cited by 5
Surf-D: Generating High-Quality Surfaces of Arbitrary Topologies Using Diffusion Models
Yu, Z., Dou, Z., Long, X., Lin, C. C., Li, Z., Liu, Y., Muller, N. E., Komura, T., Habermann, M., Theobalt, C., et al. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72933-1_24 · Cited by 25
“This approach can enhance performance without incurring additional computational costs [2,60].”
BackgroundMethodReverse Stable Diffusion: What prompt was used to generate this image?
Croitoru, F.-A., Hondru, V., Ionescu, R. T., & Shah, M. A. (2024) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2024.104210 · Cited by 15
Kim, S., Um, D., Choi, H. G., & Choi, J. Y. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3681304 · Cited by 8
Brusatin, S., Padoan, T., Coletta, A., Gatti, D. D., & Glielmo, A. (2024) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3677052.3698621 · Cited by 15
“Similarly to some existing MARL frameworks for economic modelling, we consider a sort of curriculum learning [9]: the simulation starts without the RL agents, which are introduced after 𝑡 burn-in .”
Curriculum reinforcement learning-based drifting along a general path for autonomous vehicles
Yu, K., Fu, M., Tian, X., Yang, S., & Yang, Y. (2024) · Robotica · doi:10.1017/s026357472400119x · Cited by 7
“In our specific context, we adopt a stage-based learning procedure for our agent inspired by CL [37].”
Correct like humans: Progressive learning framework for Chinese text error correction
Li, Y., Ma, S., Chen, S., Huang, H., Huang, S., Li, Y., Zheng, H.-T., & Shen, Y. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.126039 · Cited by 18
“…our proposed ProTEC framework can bring improvements to previous state-of-the-art CTEC models, thanks to the fact that it can guide CTEC models to learn and correct errors from easy to difficult like humans, which is also consistent with the core idea of widely studied curriculum learning [13].”
Peng, Z., Wang, Y., Zheng, L., et al. (2025) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2025.3541401 · Cited by 5
“Curriculum learning technique provides a promising way to alleviate the above problems [26], [27].”
Otto—Design and Control of an 8-DoF SEA-Driven Quadrupedal Robot
Scaldaferri, A., Tolomei, S., Iotti, F., Gambino, P., Pierallini, M., Angelini, F., & Garabini, M. (2025) · IEEE Open Journal of the Industrial Electronics Society · doi:10.1109/ojies.2025.3567112 · Cited by 5
“The goal of this task is to make the robot follow a user-provided desired robot base twist command Thepolicy has been trained using a game-inspired curriculum learning [27], [55], [56].”
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.