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 3,601–3,700 of 5,240
Density-based Curriculum for Multi-goal Reinforcement Learning with Sparse RewardsPreprint
Yang, D., Zhang, H., Lan, X., & Ding, J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.08903 · Cited by 5
“Our work is also highly related to curriculum learning (CL) [32], which has been introduced to deep RL to guide”
Nam, S. (2020) · Deep Blue (University of Michigan) · Cited by 2
Learning Multi-Objective Curricula for Deep Reinforcement LearningPreprint
Kang, J., Liu, M., Gupta, A., Pal, C., Liu, X., & Fu, J. (2021) · arXiv (Cornell University) · Cited by 3
On the Statistical Benefits of Curriculum LearningPreprint
Xu, Z., & Tewari, A. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.07126 · Cited by 2
“From the perspective of optimization, some papers argue that the benefits of curriculum can be interpreted as learning from more convex and more smooth objective functions, which serves as a better initialization point for the non-convex target objective function (Bengio et al., 2009).”
Contrastive Object-level Pre-training with Spatial Noise Curriculum LearningPreprint
Yang, C., Huang, L., & Crowley, E. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.13651 · Cited by 6
“Curriculum learning was first proposed in [2] to mimic the easy-to-hard learning practices that are commonly used to help with human and animal learning.”
BackgroundMethodLearning Robust Recommender from Noisy Implicit FeedbackPreprint
Wang, W., Feng, F., He, X., Nie, L., & Chua, T. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2112.01160 · Cited by 6
“Prior research on robust learning [14], [15] and curriculum learning [16] demonstrates that noisy samples are relatively harder to fit into models, exhibiting distinct patterns in the model training.”
BackgroundResultInfluentialLi, C., Zheng, P., Yin, Y., et al. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4124471 · Cited by 2
Karimi, F., Mehrpanah, A., & Rawassizadeh, R. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4531792 · Cited by 3
Tang, W. H., Sim, S. R., Aik, D. Y. K., et al. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.08.07.552352 · Cited by 2
Algorithms, Bytes, and Chips: The Emerging Political Economy of Foundation ModelsPreprint
Mills, S. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4834417 · Cited by 5
Gradual sensory maturation promotes abstract representation learningPreprint
Cheon, J., & Paik, S. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.24.661295 · Cited by 3
“A comparable strategy in machine learning is curriculum learning, where tasks are presented in order of increasing difficulty to promote generalization more effectively than na¨ıve training 48 .”
MassNet: billion-scale AI-friendly mass spectral corpus enables robust de novo peptide sequencingPreprint
Jun, A., Zhang, X., Zhang, X., Wei, J., Zhang, T.-S., Deng, Y., Liu, P., Nie, Z., Chen, Y., Dong, N., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.20.660691 · Cited by 2
Learning to see like children: proof of conceptPreprint
Gori, M., Lippi, M., Maggini, M., & Melacci, S. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1408.2478 · Cited by 2
“This resembles somehow curriculum learning [4], where examples are presented to learning systems following an increasing degree of complexity.”
BackgroundOn Simple Reactive Neural Networks for Behaviour-Based Reinforcement Learning
Pore, A., & Aragón-Camarasa, G. (2020) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra40945.2020.9197262 · Cited by 17
“However, humans tend to learn simple behaviours first in order to compose complex behaviours [18].”
BackgroundResultSupervised Image Classification with Self-Paced Regularization
Zhang, T., Gong, C., Jia, W., et al. (2018) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw.2018.00067 · Cited by 1
Modular transfer learning with transition mismatch compensation for excessive disturbance rejection
Wang, T., Lu, W., Yu, H., & Liu, D. (2022) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-022-01641-4 · Cited by 1
“In order to accelerate the learning process of RL, the knowledge previously obtained from related tasks can be used [4, 7].”
MethodEvent Nugget Detection using Pre-trained Language Models
Meghatria, R., Latiri, C., & Nader, F. (2020) · Procedia Computer Science · doi:10.1016/j.procs.2020.08.034 · Cited by 1
Trajectory-based Split Hindsight Reverse Curriculum Learning
Wu, J., Zhang, D., Zhong, S., & Qiao, H. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636842 · Cited by 2
“Curriculum learning is proposed to help agents gradually learn complex tasks by designing simple subtasks [9].”
BackgroundIntroduction to Applied Affective Computing
Tian, L., Oviatt, S. L., Muszyński, M., Chamberlain, B. C., Healey, J., & Sano, A. (2022) · ACM eBooks · doi:10.1145/3502398.3502400 · Cited by 1
Emotion Recognition in the Wild
Tian, L., Oviatt, S. L., Muszyński, M., Chamberlain, B. C., Healey, J., & Sano, A. (2022) · ACM eBooks · doi:10.1145/3502398.3502404 · Cited by 2
Learning multi-agent cooperation
Rivera, C. G., Staley, E. W., & Llorens, A. J. (2022) · Frontiers in Neurorobotics · doi:10.3389/fnbot.2022.932671 · Cited by 3
“A curriculum is anything that results in non-stationarity over the course of training (Bengio et al., 2009).”
BackgroundInformative Sample-Aware Proxy for Deep Metric Learning
Li, A., Sato, I., Ishikawa, K., Kawakami, R., & Yokota, R. (2022) · ACM International Conference on Multimedia in Asia (MMAsia) · doi:10.1145/3551626.3564942 · Cited by 1
“Pair Mining / Pair Weighting in DML Non-uniform sampling [1, 2, 14, 16] and weighting [4, 21, 22] methods have been shown to improve the performance of deep neural networks and have been widely applied to various tasks.”
BackgroundCurriculum Defense: An Effective Adversarial Training Method
Yin, H., Deng, X., & Yan, J. (2022) · Chinese Control Conference (CCC) · doi:10.23919/ccc55666.2022.9901570 · Cited by 1
“Curriculum learning [10] can be seen as such a del-This work is supported by National Natural Science Foundation (NNSF) of China under Grant 62133011. icate mechanism in the family of machine learning.”
BackgroundInfluentialHayashi, S. Y., & Hirata, N. S. T. (2022) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr56361.2022.9956133 · Cited by 1
“This type of error analysis hints to the possibility of employing curriculum learning approaches [20], in an incremental fashion.”
BackgroundFast Radio Propagation Prediction with Deep Learning
Levie, R., Yapar, Ç., Caire, G., & Kutyniok, G. (2022) · Applied and numerical harmonic analysis · doi:10.1007/978-3-031-09745-4_10 · Cited by 2
Richer, T. J., & Standen, M. (2023) · Advances in information security · doi:10.1007/978-3-031-29269-9_13 · Cited by 1
The Information Pathways Hypothesis: Transformers are Dynamic Self-Ensembles
Hussain, M. S., Zaki, M. J., & Subramanian, D. (2023) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3580305.3599520 · Cited by 5
“This method falls within the category of curriculum learning [6] strategies such as [36, 47].”
MethodAutonomous Stair Ascending and Descending by Quadruped Wheelchairs
Akamisaka, A., & Nagao, K. (2023) · International Conference on Ubiquitous Robots (UR) · doi:10.1109/ur57808.2023.10202377 · Cited by 1
“Curriculum Learning Curriculum learning [29] is a machine learning approach that elaborates learning procedures to improve learning speed and accuracy.”
BackgroundA multi-domain adaptive neural machine translation method based on domain data balancer
Xu, J., Wen, Y., Huang, S., & Yu, Z. (2023) · Intelligent Data Analysis · doi:10.3233/ida-230155 · Cited by 1
“In Data-Centric Adaptation methods, Zhang et al. [55] attempted to apply curriculum learning [6] to domain adaptation, gradually transitioning unlabeled network data to the corresponding target domain through curriculum design to expand target domain data.”
A Model-based Curriculum Learning Strategy for Training SegFormer
Cambrin, D. R., Apiletti, D., & Garza, P. (2023) · IEEE International Conference on Application of Information and Communication Technologies (AICT) · doi:10.1109/aict59525.2023.10313143 · Cited by 1
“Many different criteria were used for evaluating the difficulty of the samples, ranging from simple shape analysis [13] to more complex factors such as the presence of fog [10].”
BackgroundMethodA Taxonomy for Learning with Perturbation and Algorithms
Yao, R., & Wu, O. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3644391 · Cited by 1
Histopathology Image Classification With Noisy Labels via The Ranking Margins
Wen, Z., Wu, H., & Ying, S. (2024) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2024.3381775 · Cited by 2
“In the early studies, [42] and [43], curriculum learning was pre-determined and fixed during the training stage.”
BackgroundAppearance-Based Curriculum for Semi-Supervised Learning with Multi-Angle Unlabeled Data
Tanaka, Y., Yoshida, S. M., Shibata, T., Terao, M., Okatani, T., & Sugiyama, M. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00276 · Cited by 1
Instance-Ambiguity Weighting for Multi-label Recognition with Limited Annotations
Shrewsbury, D., Kim, S., Kim, Y.-E., Kong, H., & Lee, S. (2024) · Lecture notes in computer science · doi:10.1007/978-981-97-2242-6_13 · Cited by 2
Fault Classification Method based on Self-paced Neighborhood Preserving Embedding
Gao, H., Ma, C., & Zhang, Y. (2024) · International Symposium on Autonomous Systems (ISAS) · doi:10.1109/isas61044.2024.10552614 · Cited by 1
“S ELF - PACED L EARNING Curriculum Learning [16] was proposed by the team of Yoshua Bengio.”
MethodGuo, J., Kang, Y., Li, X., Zhang, W., Kuang, K., Sun, C., Tang, S., & Wu, F. Q. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3417235 · Cited by 3
“Comparedtotraditionalcurriculumlearning’spre-determined heuristiccurriculum[35],[36],SPLallowsthemachineitself to automatically set the difficulty and sequence of curriculum according to the progress of learning and the cognitive level of the model.”
BackgroundCurriculum Learning Driven Domain Adaptation for Low-Resource Machine Reading Comprehension
Zhang, L., Wang, Q., Xu, B., Liu, Y., & Mao, Z. (2024) · IEEE Signal Processing Letters · doi:10.1109/lsp.2024.3429270 · Cited by 1
“For curriculum arrangement, we adopt Naive Order and Baby Step [15] bucket will be merged into the training subset after several steps, as shown in Fig.”
BackgroundMethodOnline Optimization of Curriculum Learning Schedules using Evolutionary Optimization
Jiwatode, M., Schlecht, L., & Dockhorn, A. (2024) · IEEE Conference on Games (CoG) · doi:10.1109/cog60054.2024.10645570 · Cited by 1
“Bengio et al. [4] defined CL as a training strategy for ML algorithms using a curriculum of tasks.”
MethodCross-Domain Document Summarization Model via Two-Stage Curriculum Learning
Lee, S., Kim, G., & Kang, S. (2024) · Electronics · doi:10.3390/electronics13173425 · Cited by 1
Learning with Noisy Labels: A Novel Meta Label Corrector without Clean Validation Data
Jiang, K. Y., Cao, B., Liu, Y., & Fan, J. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10651351 · Cited by 2
“Especially in the initial stages of training, the model is anti-noise that it tends to learn clean data more than noisy data [28], [29].”
BackgroundImportance-Guided Sequential Training for Physics-Informed Neural Networks
Zhang, X., Chen, N., Qiu, J., Shi, P., Wu, X., & Yuan, W. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10651329 · Cited by 1
“A nat-ural approach inspired by curriculum learning [35] involves learning complex tasks in multiple steps.”
MethodMobile Robots through Task-Based Human Instructions using Incremental Curriculum Learning
Muttaqien, M. A., Yorozu, A., & Ohya, A. (2024) · IEEE Conference on Cybernetics and Intelligent Systems · doi:10.1109/cis-ram61939.2024.10673231 · Cited by 2
“In this paper, we delve into the domain of task-based human instructions and their integration with curriculum learning strategies [3] [4] to empower mobile robots with enhanced navigational capabilities in complex indoor environments.”
BackgroundAdvanced fine-tuning procedures to enhance DNN robustness in visual coding for machines
Marie, A., Desnos, K., Mercat, A., Morin, L., Vanne, J., & Zhang, L. (2024) · EURASIP Journal on Image and Video Processing · doi:10.1186/s13640-024-00650-3 · Cited by 1
“This procedure, first introduced in our previous work [16], shares some similarities with curriculum learning training procedures [32, 33].”
ResultLarge Language Model for Action Anticipation
Li, W., Luo, D., Yang, D., & Wang, W. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72338-4_15 · Cited by 2
Curriculum Learning for Multimedia in the Era of Large Language Models
Wang, X. Y., Zhou, Y., Chen, H., & Zhu, W. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3689178 · Cited by 1
“Curriculum learning (CL) has continuously gained attention since its first advent in 2009 [1].”
Liu, T., Lam, K., & Bao, B.-K. (2024) · International Journal of Computer Vision · doi:10.1007/s11263-024-02279-1 · Cited by 2
“…different from the common practice that directly trains the online detector to anticipate the future semantics based on historical observations, we leverage a curriculum learning (Bengio et al., 2009; Soviany et al., 2022) paradigm to seek the assistance of auxiliary information from future frames.”
Multi-agent Reinforcement Learning for Cooperative On-Ramp Merging of Connected Automated Vehicles
Zhao, B., & Cui, J. (2024) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-031-70507-6_16 · Cited by 3
Enhancing Spatio-temporal Quantile Forecasting with Curriculum Learning: Lessons Learned
Yin, D., Deng, J., Ao, S., Li, Z., Xue, H., Prabowo, A., Jiang, R., Song, X., & Salim, F. D. (2024) · ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL GIS) · doi:10.1145/3678717.3691216 · Cited by 1
“The concept of curriculum learning (CL) emerges as a promising option to achieve high-quality and efficient training, where the constructed model is trained on instances of increasing difficulty, mimicking the learning process of human beings [2].”
Choi, S. H., Oh, J., & Cha, J.-W. (2024) · AI · doi:10.3390/ai5040133 · Cited by 1
Tai, T. C., Do-Tran, N.-T., Le, N.-H.-L., Li, Y., & Huang, C.-C. (2024) · Lecture notes in computer science · doi:10.1007/978-981-96-0969-7_1 · Cited by 1
“We also incorporate curriculum learning [1] techniques, allowing the model to progress from simple to complex samples.”
DCLNet: Data Closed-Loop Network for Laryngoscopy Image Annotation and Classification
Xu, X., Zhao, X., Cao, J., Xu, W., Fu, T., Ren, R., Hu, Z., & Zhang, X. (2024) · IEEE Biomedical Circuits and Systems Conference (BioCAS) · doi:10.1109/biocas61083.2024.10798144 · Cited by 1
“Curriculum Learning [26]-[28] is a machine learning paradigm that draws inspiration from educational learning strategies.”
A contrastive news recommendation framework based on curriculum learning
Zhou, X., Lin, N., Zheng, W., Zhou, D., & Yang, A. (2024) · User Modeling and User-Adapted Interaction · doi:10.1007/s11257-024-09422-0 · Cited by 3
“The concept of curriculum learning (Bengio et al. 2009) was initially introduced by Bengio et al., who pointed out that by gradually introducing more complex training data, deep neural networks can be helped to converge faster and achieve higher training performance.”
FeedMatch: Evolves for Semi-Supervised Multimedia Classification from Student Feedback
Liu, J., Sun, D., Xia, H., Bai, J., & Fan, X. (2024) · ACM International Conference on Multimedia in Asia (MMAsia) · doi:10.1145/3696409.3700218 · Cited by 1
“Curriculum learning [3] has been utilized to improve the model’s ability to balance class-wise learning.”
Single-cell Curriculum Learning-based Deep Graph Embedding Clustering
Li, H., Fu, J., Ling, X., Sun, Z., Wang, K., & Chen, Z. (2024) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm62325.2024.10822530 · Cited by 1
“In particular, curriculum learning (CL) is an effective training strategy for gradually guiding model learning in tasks with obvious difficulty levels [17].”
Baldão, T. P., Máximo, M. R. O. A., & Yoneyama, T. (2024) · Brazilian Symposium on Robotics and Workshop on Robotics in Education (SBR/WRE) · doi:10.1109/sbr/wre63066.2024.10837743 · Cited by 2
Continual debiasing: A bias mitigation framework for natural language understanding systems
Lee, M., Kim, J., Park, J.-H., & Lee, S. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.126593 · Cited by 2
Meta-reinforcement Learning Task Planning Based on Improved Curriculum Learning Sampling
Zhang, W., Wang, J., & Wan, N. (2024) · China Automation Congress (CAC) · doi:10.1109/cac63892.2024.10864762 · Cited by 1
F-StrIPE: Fast Structure-Informed Positional Encoding for Symbolic Music Generation
Agarwal, M., Wang, C., & Gaël Richard, A. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10888904 · Cited by 1
“Training for 15 epochs with a batch size of 8, we use gradient clipping and curriculum learning [22].”
PHFS: Progressive Hierarchical Feature Selection Based on Adaptive Sample Weighting
Zhao, H., Shi, J., & Zhang, Y. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3525643 · Cited by 4
“Curriculum learning, as a representative of fixed selection strategies, has become a research hotspot due to its denoising and guidance functions [22].”
PatchST: A Patch Spatial-Temporal Network for Large-Scale Traffic Forecasting
Li, J., Zhang, G., Wang, W., et al. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10890862 · Cited by 1
“Taking inspiration from MTGNN [9], we incorporate a training strategy called curriculum learning, a concept initially proposed by Bengio [24].”
Fu, P., & Guo, S. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10888178 · Cited by 1
“To address this, we propose a curriculum learning (CL) strategy inspired by human learning [26], progressing from easier to harder lesion types.”
Bae, J., & Shin, J. (2025) · International Conference on Artificial Intelligence in Information and Communication (ICAIIC) · doi:10.1109/icaiic64266.2025.10920651 · Cited by 1
“Curriculum learning [15] is a strategy that structures the training process into sequential phases based on specific objectives.”
Duminil, A., Ieng, S.-S., & Gruyer, D. (2024) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc58415.2024.10919905 · Cited by 1
“However, in order to include adverse weather condition new neural network architecture based on both Xception [37] and curriculum learning approach [38] is used.”
Self-training dual-network for denoising federated recommendation
Liu, P., Lu, G., Lv, S., & Wang, Z. (2025) · Journal of Intelligent Information Systems · doi:10.1007/s10844-025-00930-y · Cited by 1
“Generally, we can observe that: (1) The average predicted probabilities for noisy samples are lower than those for clean samples after several rounds, aligning with the expectations from curriculum learning research (Bengio et al., 2009; Han et al., 2018).”
InfluentialA Deep Reinforcement Learning Tracking Algorithm Based on Task Decomposition
Yang, K., Shen, A., Xu, N., et al. (2024) · International Conference on Industrial Automation, Robotics and Control Engineering (IARCE) · doi:10.1109/iarce64300.2024.00043 · Cited by 1
“However, curriculum learning creates a curriculum to allow the agent to use knowledge acquired from simpler tasks to tackle more challenging ones [11].”
DimCL: Dimension-Aware Augmentation in Contrastive Learning for Recommendation
Zhang, C., Han, Q., Tan, Q., Wang, S., Zhao, X., & Chen, R. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3690624.3709200 · Cited by 14
Xu, H., Chen, H., Liu, X., & Ren, H. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.110622 · Cited by 2
Learning Keypoints for Multi-Agent Behavior Analysis using Self-Supervision
Khalil, D., Liu, C., Perona, P., Sun, J. J., & Marks, M. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00066 · Cited by 2
“Our final loss function for B-KinD-multi combines the reconstruction loss L recon , rotation equivariance loss L r , and separation loss L s : We employ curriculum learning [5], applying L r and L s only after the keypoints are consistently discovered from the semantic parts of each agent instance.”
Bayesian curriculum generation in sparse reward reinforcement learning environments
Akgün, O., & Üre, N. K. (2025) · Engineering Science and Technology an International Journal · doi:10.1016/j.jestch.2025.102048 · Cited by 1
REED‐VAE: RE‐Encode Decode Training for Iterative Image Editing with Diffusion Models
Almog, G., Shamir, A., & Fried, O. (2025) · Computer Graphics Forum · doi:10.1111/cgf.70020 · Cited by 1
“Effectively, we perform a variation of curriculum learning [BLCW09], which attempts to mimic human learning by gradually increasing the complexity of data samples used when training a model.”
UPCMR: A Universal Prompt-Guided Model for Random Sampling Cardiac MRI Reconstruction
Lyu, D., Rao, C., Staring, M., van Osch, M. J. P., Doneva, M., Lamb, H. J., & Pezzotti, N. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-87756-8_44 · Cited by 4
Li, Y., Yu, F., & Chen, Q. (2025) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2025.3566923 · Cited by 2
“Bengio et al. [18] noted that students must typically master foundational subjects prior to undertaking more complex subjects, leading to the development of the concept known as course learning.”
A Model-Agnostic Hierarchical Framework Towards Trajectory Prediction
Qian, T., Wang, Y., Xu, Y., Zhang, Z., Wu, L., Qiu, Q., & Wang, F. (2025) · Journal of Computer Science and Technology · doi:10.1007/s11390-023-3013-4 · Cited by 2
Chinnaraju, A. (2024) · Journal of Emerging Technologies and Innovative Research · doi:10.56975/jetir.v11i11.559859 · Cited by 4
Tao, H., Jia, Z., Zhu, B., & Lin, T.-Y. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.110948 · Cited by 1
RetractedCurriculum Learning for Few-Shot Domain Adaptation in CT-Based Airway Tree Segmentation
Jacovella, M., Keshavarzi, A., & Angelini, E. D. (2025) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi60581.2025.10980795 · Cited by 1
“Originally formalized by Bengio et al. in 2009 [12], CL is about “starting small”, by forcing a model to first focus on simpler parts of the learning task, or a simplified sub-task, before progressively increasing difficulty.”
Chang, H., Gu, L., Hu, C., Zhang, Z., Zhu, H. C., Xu, Y., Fang, Y., & Chen, Z. (2025) · International World Wide Web Conference (WWW) · doi:10.1145/3701716.3715260 · Cited by 2
Solving the forward problem of the Falkner–Skan equation using physics-informed derivative networks
Cong, K., Li, G., Sun, Y., & Ren, H. (2025) · Zeitschrift für Naturforschung A · doi:10.1515/zna-2025-0128 · Cited by 2
Efficient neural network training method for unsteady flow field prediction based on data pool
Min, Q., Li, T., Li, G., et al. (2025) · Engineering Applications of Computational Fluid Mechanics · doi:10.1080/19942060.2025.2512958 · Cited by 1
“To improve the neural network’s accu racy, we designed and implemented a curriculum learning-based time-step reset mechanism, which guides the neural network to progressively learn from simple to complex tasks (Bengio et al., 2009).”
SASFNet: Soft-edge awareness and spatial-attention feedback deep network for blind image deblurring
Cheng, J., Zhang, K., Hou, J., Zhang, Y., & Shi, G. (2025) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2025.104408 · Cited by 1
CLCoSum: Curriculum Learning-Based Code Summarization for Code Language Models
He, H., Wang, J., Cao, L., & Cai, Y. (2025) · IEEE International Conference on Program Comprehension (ICPC) · doi:10.1109/icpc66645.2025.00035 · Cited by 1
“To address this issue, we propose leveraging Curriculum Learning (CL) [23], inspired by the natural progression from easy to hard in human learning, to aid CLMs adapt more effectively to challenging samples.”
Joint Imbalance Adaptation for Radiology Report Generation
Li, W., Han, G., Wu, Y., Huang, I., & Huang, X. (2025) · Journal of Healthcare Informatics Research · doi:10.1007/s41666-025-00205-9 · Cited by 3
Bioinspired morphology and task curricula for learning locomotion in bipedal muscle-actuated systems
Badie, N., Al-Hafez, F., Schumacher, P., Haeufle, D. F. B., Peters, J., & Schmitt, S. (2025) · Communications Engineering · doi:10.1038/s44172-025-00443-0 · Cited by 4
“Curricula break downcomplextasksintosimplerstages,facilitatinglearningthroughguided progression 21 .”
Catheter detection and segmentation in X-ray images via multi-task learning
Lin, X., Ma, Y., Koland, E., Howell, S., Rinaldi, C. A., & Rhode, K. (2025) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-025-03461-7 · Cited by 3
“Inspired by CL [22, 25], our model is designed to prioritize difficult samples and tasks simultaneously, similar to the optimization of hard negative examples [27, 28] across different samples and tasks.”
Shaping Reward Signals in Reinforcement Learning Using Constraint Programming
Yin, C. N., Cappart, Q., & Pesant, G. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-95976-9_16 · Cited by 1
Semi-Supervised Maritime Object Detection: A Data-Centric Perspective
Wu, M., Zhang, W., Min, R., Zhang, L., Xu, Y., Qin, Y., & Yu, J. J. (2025) · Journal of Marine Science and Engineering · doi:10.3390/jmse13071242 · Cited by 2
UAV Swarm Confrontation Decision Optimization Based on MADDPG
Liu, Y., Li, D., & Meng, Z. (2025) · doi:10.1109/icaisisas64483.2025.11051494 · Cited by 1
Han, D., Kim, B., Lee, R., Kim, K., Hwang, H., & Kang, H. (2025) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/3721238.3730738 · Cited by 2
“Curriculum learning [Bengio et al. 2009] organizes training tasks in increasing order of difficulty, a strategy shown to accelerate convergence and improve performance in deep learning [Baker et al. 2019; Hacohen and Weinshall 2019; Weinshall et al. 2018].”
Automated Coding Utterances Toward Chinese Course Core Competence with Large Language Models
Yue, M., Qi, T., Wang, M., Ruan, W., He, J. Y. D., Sun, B., & Zheng, G. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-0014-7_19 · Cited by 1
Proxy-Validated Importance-Aware Federated Sample Selection with Meta Learning
Zhang, Y., Miao, X., Li, B., Wu, Y., & Shang, Y. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3711896.3737093 · Cited by 4
“Another line of work dynamically selects data during the training process, including importance sampling-based methods [11, 39], curriculum learning-based methods [2], and reinforcement learning-based methods [42].”
Wang, Z., Zhao, Y., Gong, M., Li, H., Liu, G., & Jian-long, T. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111870 · Cited by 1
RAFT: Regularized Adversarial Fine-Tuning to Enhance Deep Reinforcement Learning for Self-Parking
Pighetti, A., Bellotti, F., Berta, R., Cavallaro, A., Lazzaroni, L., & Oh, C. (2025) · IEEE Sensors Letters · doi:10.1109/lsens.2025.3600982 · Cited by 2
“However, DRL agents are difficult to train (see, e.g., [3]), in which issue is addressed by several techniques (e.g., curriculum learning [4], [5]).”
Xue, S., Yuan, L., Lv, K., Ran, T., Xiao, W., Zhang, J., Wang, G., & Cui, J. (2025) · Control Engineering Practice · doi:10.1016/j.conengprac.2025.106538 · Cited by 2
IsGCL: Informative Sample-Aware Progressive Graph Contrastive Learning
Zeng, J., Wang, P., Ma, L., Tao, J., & Guan, X. (2025) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde65448.2025.00137 · Cited by 3
“Instead, inspired by curriculum learning mechanism [7], [19], [20], whose basic insight is to imitate the learning mechanism of humans and train models from easier to harder views, we propose a progressive confident views selection strategy to sample a small amount of highly confident views…”
Synthesizing Evolving Symbolic Representations for Autonomous Systems
Sartor, G., Oddi, A., Rasconi, R., Santucci, V. G., & Meo, R. (2025) · Progress in Artificial Intelligence · doi:10.1007/s13748-025-00394-9 · Cited by 1
“This knowledge can later be leveraged to solve user-assigned tasks [31] or serve as a scaffold for acquiring further knowledge incrementally [32–34], as in curriculum learning [35]).”
Li, Y., Ma, Y., Shen, X., & Liu, C. (2025) · Multimedia Systems · doi:10.1007/s00530-025-01972-6 · Cited by 1
“A typical Curriculum Learning strategy consists of two core components: a dif - ficulty measure and a training scheduler[ 41 ].”
Tian, F., Liu, Z., Popov, S., et al. (2025) · Journal of Lightwave Technology · doi:10.1109/jlt.2025.3602509 · Cited by 1
“This adaptive loss weighting strategy can be regarded as a training heuristic, similar to annealed objective weighting or curriculum learning techniques commonly employed in neural network optimization [34], [35].”
Comparative Benchmark of Sampling-Based and DRL Motion Planning Methods for Industrial Robotic Arms
Astorquia, I. F., Villate-Castillo, G., Tellaeche, A., & Vázquez, J. I. (2025) · Sensors · doi:10.3390/s25175282 · Cited by 5
“The agent was trained in a simulated environment using a hybrid reward function, expert demonstrations, and curriculum learning [65] to progressively acquire high-precision motion skills across the robot’s full workspace.”
Nakaya, R., Harada, T., Miura, Y., Hattori, K., & Matsuoka, J. (2025) · Artificial Life and Robotics · doi:10.1007/s10015-025-01056-8 · Cited by 2
“Inspired by curriculum learning [4], we progressively increase task complexity by decreasing this interval over the course of training.”
Dynamic Environment Adaptation in Biped Robots: A Reinforcement Learning Approach
Soto, M., Ormeño-Arriagada, P., Vásquez, J., et al. (2025) · IEEE Access · doi:10.1109/access.2025.3604236 · Cited by 1
“Subsequently, the performance of the models is reported within a curriculum learning framework [45], progressively increasing task difficulty.”
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.