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 2,801–2,900 of 5,240
Bian, S., Chen, X., Zhao, W. X., et al. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.13299 · Cited by 5
“with noisy labels. Basically, learning with noisy labels aims to solve the problem when some of the training data is unreliable in supervised learning. Early methods, such as the curriculum learning [1] and self-paced learning [16], alleviate this problem by reordering the training samples. Easy instances will be lear…”
BackgroundLifelong Language Knowledge DistillationPreprint
Chuang, Y.-S., Su, S., & Chen, Y.-N. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.02123 · Cited by 5
“In the right-to-left order, the baseline LAMOL can easily outperform multi-task models due to its easiest-to-hardest order, which helps the model to better transfer the knowledge gradually in these NLG tasks, similar to curriculum learning (Bengio et al., 2009).”
MethodTraining Stronger Baselines for Learning to OptimizePreprint
Chen, T., Zhang, W., Zhou, J., Chang, S., Liu, S., Amini, L. D., & Wang, Z. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.09089 · Cited by 37
“The idea [27] is to first focuses on learning from a subset of simple training examples, and gradually expanding to include the remaining harder samples.”
BackgroundMethodInfluentialMeta Automatic Curriculum LearningPreprint
Portelas, R., Romac, C., Hofmann, K., & Oudeyer, P. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.08463 · Cited by 11
“The idea of organizing the learning sequence of a machine is an old concept that stems from multiple works in reinforcement learning [1, 2], developmental robotics [3] and supervised learning [4, 5], from which the Deep RL community borrowed the term Curriculum Learning.”
MethodZero-Shot Learning with Knowledge Enhanced Visual Semantic EmbeddingsPreprint
Sikka, K., Huang, J., Silberfarb, A., Nayak, P., Rohrer, L., Sahu, P., BYRNES, J. J., Divakaran, A., & Rohwer, R. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.10889 · Cited by 6
“This is similar to using curriculum learning [8], but without the need for manual creation of a curriculum.”
MethodSAfE: Self-Attention Based Unsupervised Road Safety Classification in Hazardous Environments.Preprint
Kothandaraman, D., Chandra, R., & Manocha, D. (2020) · arXiv (Cornell University) · Cited by 4
Revisiting Iterative Back-Translation from the Perspective of Compositional GeneralizationPreprint
Guo, Y.-N., Zhu, H., Lin, Z., Chen, B., Lou, J.-G., & Zhang, D.-M. (2020) · AAAI Conference on Artificial Intelligence · doi:10.48550/arxiv.2012.04276 · Cited by 30
“Towards this end, we consider curriculum learning (Bengio et al. 2009) as a potential solution: start out iterative back-translation with easy monolingual data (of which the generated pseudo-parallel data are less error-prone), then gradually increase the difficulty of monolingual data during the training process.”
BackgroundMeta-Learning Bandit Policies by Gradient AscentPreprint
Kveton, B., Mladenov, M., Hsu, C., Zaheer, M., Szepesvári, C., & Boutilier, C. E. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.05094 · Cited by 20
Uncertainty-Aware Unsupervised Domain Adaptation in Object DetectionPreprint
Guan, D., Huang, J., Xiao, A., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.00236 · Cited by 5
“Curriculum learning has been widely explored in the past decade [55]–[57].”
Contrastive Neural Architecture Search with Neural Architecture ComparatorsPreprint
Chen, Y., Guo, Y. F., Chen, Q., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.05471 · Cited by 5
“Moreover, we develop a data exploration method that trains NAC in a semi-supervised way to reduce the requirement of the training data.”
MethodAutoLoss: Automated Loss Function Search in RecommendationsPreprint
Zhao, X., Liu, H., Fan, W., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.06713 · Cited by 5
“Motivated by curriculum learning [1, 19], the original class probabilities should be generated according to the ground truth labels 𝒚 and the output of DRS network ˆ 𝒚 .”
BackgroundImproving Sentence-Level Relation Extraction through Curriculum LearningPreprint
Park, S., & Kim, H. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.09332 · Cited by 12
“In the curriculum learning process, general parameters for relation extraction are learned through easy examples and high-level inference becomes possible while difficult examples are gradually learned [3].”
BackgroundRule-Based Interactive Assisted Reinforcement Learning
Bignold, A., Vamplew, A. /. P., Dazeley, A. /. R., & Foale, C. (2019) · FedUni ResearchOnline (Federation University Australia) · doi:10.13140/rg.2.2.27422.48962 · Cited by 4
Cai, T., Xie, L., Chen, M., Liu, Y., He, D., Zhang, S., Mura, C., Bourne, P. E., & Xie, L. (2021) · Research Square · doi:10.21203/rs.3.rs-1109318/v1 · Cited by 6
“Curriculum learning [61] emphasises to sort all training so that the model is exposed to challenges of increasing difficulty.”
S2match: Self-Paced Sampling for Data-Limited Semi-Supervised LearningPreprint
Guan, D., Xing, Y., Huang, J., et al. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4634958 · Cited by 5
Processing and learning deep neural networks on chip
Hacene, G. B. (2019) · theses.fr (ABES) · doi:10.70675/31922a0ezdf1bz462dz8909z687397ad8117 · Cited by 8
C-KGE: Curriculum learning-based Knowledge Graph Embedding
Zhou, D., Li, S., Dong, L., Chen, R., Peng, X., & Yao, H. L. (2024) · Computer Speech & Language · doi:10.1016/j.csl.2024.101689 · Cited by 5
Balanced Domain Randomization for Safe Reinforcement Learning
Kang, C., Chang, W., & Choi, J. (2024) · Applied Sciences · doi:10.3390/app14219710 · Cited by 7
FedSeq: Personalized Federated Learning via Sequential Layer Expansion in Representation Learning
Jang, J. W., & Choi, B. J. (2024) · Applied Sciences · doi:10.3390/app142412024 · Cited by 5
Zhang, T.-Y., Yan, Z., Li, C. X., Ying, N., Lei, Y., Lyu, S., Feng, Y., Zhao, Y., & Zhang, G. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3554752 · Cited by 9
“In 2009, Bengio et al. [23] proposed a learning paradigm called CL.”
Toward Open-World Domain Adaptation via Iteratively Contrastive Learning and Clustering
Li, J., Sun, H., Li, J., Chen, P., & Wei, S. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3584072 · Cited by 5
“Inspired by the curriculum learning [71], we first consider partial instances with reliable cluster assignments, and as training proceeds, more reliable instances are considered.”
Tang, K., Huang, X., Ren, F., et al. (2025) · Computers & Structures · doi:10.1016/j.compstruc.2025.107937 · Cited by 4
Deep neural network based precise orbit prediction for low earth orbit (LEO) satellites
Li, B., Wu, T., & Ge, H. (2025) · Journal of Geodesy · doi:10.1007/s00190-025-01999-7 · Cited by 8
“In addition, some training strategies, such as curriculum learning (Bengio et al. 2009) are applied to ensure a better model performance on the test dataset.”
Intrinsic Decomposition-Based Curriculum Learning for Low-Light Image Enhancement
Kim, J., Park, G., Bae, J., & Jung, S. (2025) · IEEE Signal Processing Letters · doi:10.1109/lsp.2025.3621115 · Cited by 7
Vision-Based Reinforcement Learning for Robotic Grasping of Moving Objects on a Conveyor
Cao, Y., Xu, X., & Zhang, Y. (2025) · Machines · doi:10.3390/machines13100973 · Cited by 7
“To ensure a fair comparison among reinforcement learning algorithms, we implemented a curriculum learning strategy [30–32], which progressively increased the task difficulty.”
An Information Theory of Compute-Optimal Size Scaling, Emergence, and Plateaus in Language Models
Nayak, A., & Varshney, L. (2024) · IEEE Journal on Selected Topics in Signal Processing · doi:10.1109/jstsp.2025.3626264 · Cited by 7
“Curriculum learning results [22] further show that models tend to acquire simpler patterns before more complex ones, suggesting a sequential progression in learned abstractions similar to the peeling-based learning [17].”
Recent Advances in Deep Domain Adaptation Research for Semantic Segmentation in Urban Scenes
Zhu, S., Tai, Q., Du, L., Miao, L., Liu, X., Li, N., & Hou, S. (2025) · Mathematics · doi:10.3390/math13223611 · Cited by 4
Transformer-based sequence-to-sequence soft sensor using missing data in industrial processes
Liu, Y., Gao, J., Shao, J., Wu, J., Jia, M., & Liu, Q. (2026) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/j.chemolab.2026.105631 · Cited by 6
The Internet of Humanoids: A Survey of Technologies, Applications, and Challenges
Yu, A. W., & Nayak, A. R. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3653698 · Cited by 9
“To address learning e ffi ciency, approaches such as o ffl ine RL [37], model-based RL [38], and curriculum learning [39] are increasingly explored.”
Fu, D., Deng, Y., Wang, H., et al. (2026) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2026.114232 · Cited by 4
Harnessing Massive Satellite Imagery with Efficient Masked Image Modeling
Wang, F., Wang, H., Wang, D., Guo, Z., Zhong, Z., Lan, L., Yang, W., & Zhang, J. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00652 · Cited by 46
Li, M., Zhou, T., Han, B., Liu, T., Liang, X., Zhao, J., & Gong, C. (2024) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2024.3377123 · Cited by 6
“Multi-Task Curriculum learning Framework (MTCF) [23] utilizes curriculum learning [41], [42] to construct an ordered sequence, and weighs ID examples and OOD examples differently from easy to difficult.”
MethodResearch on UAV Coverage Search Based on DDQN in Unknown Environments
Deng, G., Yao, X., Wang, B., He, X. S., & Fei, Q. (2023) · China Automation Congress (CAC) · doi:10.1109/cac59555.2023.10451197 · Cited by 4
Dynamic curriculum learning for conversation response selection
Chen, G., Zhan, R., Wong, D. F., & Chao, L. S. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.111687 · Cited by 4
Survival games for humans and machines
Strannegård, C., Engsner, N., Ulfsbäcker, S., Andreasson, S., Endler, J. A., & Nordgren, A. (2024) · Cognitive Systems Research · doi:10.1016/j.cogsys.2024.101235 · Cited by 5
“These deep RL algorithms can be combined with curriculum learning (Bengio, Louradour, Collobert, & Weston, 2009), which divides the learning process into stages of increasingly difficult environments; continual learning , with agents that continue to learn throughout their lifetime (Khetarpal,…”
Liu, K., Zhou, Y., & Huang, H. (2024) · Complex & Intelligent Systems · doi:10.1007/s40747-024-01485-1 · Cited by 7
“Curriculum learning (CL), as advocated by Bengio [40], suggests that models should initially learn from easier samples and gradually transition to more complex ones.”
Xu, Q., Liu, L., & Zhou, X. (2024) · IEEE International Joint Conference on Neural Network · doi:10.1109/ijcnn60899.2024.10651238 · Cited by 4
“• We introduce the concept of curriculum learning [12], [13], and establish global and local confidence thresholds that are dynamically adjusted during training to balance the quantity and quality of pseudo-labels.”
BackgroundDifferentiable Micro-Mesh Construction
Dou, Y., Zhong, Z., Jin, Q., Shi, R., Li, Y., & Ni, B. (2024) · 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/cvpr52733.2024.00411 · Cited by 5
“The above derivation are formulated in the level of detail context intentionally, as we design the training procedure under a curriculum learning paradigm [4], shown in next: 1. ℓ = 0 : Warmup with optimizing base mesh, which is similar to the mesh deformation under fixed topology.”
Domain Adaptation Transformer for Unsupervised Driving-Scene Segmentation in Adverse Conditions
Liu, W., Wang, S., Zhu, J., Xie, X., & Zhang, L. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3461468 · Cited by 8
Navigating Beyond Instructions: Vision-and-Language Navigation in Obstructed Environments
Hong, H., Wang, S., Huang, Z. H., Wu, Q., & Liu, J. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3681640 · Cited by 17
“To address this, we develop the Obstructed VLN (ObVLN) method, including a curriculum training strategy [8] to organize the training and a novel graph construction mechanism to introduce virtual nodes for blocked edges to facilitate efficient exploration.”
MethodWang, H., Guo, J., Zhang, B., Wang, Z., Li, X., & Tang, S. (2024) · Journal of Aerospace Engineering · doi:10.1061/jaeeez.aseng-5746 · Cited by 4
Kim, H., Choi, J., Do, H., & Lee, G. T. (2024) · Drones · doi:10.3390/drones9010026 · Cited by 6
A hybrid curriculum learning and tree search approach for network topology control
Meppelink, G. J., Rajaei, A., & Cremer, J. L. (2025) · Electric Power Systems Research · doi:10.1016/j.epsr.2025.111455 · Cited by 8
Teutsch, P., Pfeffer, P., Ghazijahani, M. S., et al. (2025) · APL Machine Learning · doi:10.1063/5.0244416 · Cited by 4
“Curriculum learning strategies alter the task for the model over the course of training epochs which can stabilize training and speed up convergence 45 .”
A Survey of Curriculum Learning in Deep Reinforcement Learning
Lin, S., Mi, Q., & Gao, T. (2025) · IEEE Annual Computing and Communication Workshop and Conference (CCWC) · doi:10.1109/ccwc62904.2025.10903795 · Cited by 13
“…the concept of curriculum learning (CL), which aims to control the learning trajectory (i.e. sequence) of the agent, thereby enhancing either the agent's capacity to acquire samples or its utilization of samples, ultimately improving sample efficiency to accelerating the training process [3].”
A Noisy Label Filter based on GMM Binary Classification for Speaker Verification
Yao, J., Lu, S., Ren, P., Wang, K., Hong, Q., & Li, L. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10889777 · Cited by 6
“As the training continues, simple samples are selected first, followed by increasingly hard samples, aligning with the concept of curriculum learning [22], [23].”
Leveraging LLM and Text-Queried Separation for Noise-Robust Sound Event Detection
Yin, H., Xiao, Y., Bai, J., & Das, R. K. (2025) · IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) · doi:10.1109/icasspw65056.2025.11011161 · Cited by 8
Chong, C. F., Fang, X., Guo, J., Abreu, P. H., Wang, Y., Yang, X., Ke, W., & Im, S. (2025) · Neurocomputing · doi:10.1016/j.neucom.2025.130498 · Cited by 7
“Durand et al. [7] proposed a curriculum learning-based method [41] with a graph neural network that explicitly models the correlations between categories to generate pseudo-labels gradually.”
Strand, A., Gorton, P. R., & Bråthen, K. (2025) · IEEE Access · doi:10.1109/access.2025.3580279 · Cited by 22
“Curriculum learning can be seen as a special form of transfer learning in which initial tasks are used to guide the learning process and improve performance on the final task [56].”
Autonomous Driving: Integration of Segmentation and Depth Camera in a Curriculum Learning Approach
Barra, S., Cimmino, L., Loia, V., Nappi, M., & Polsinelli, M. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3579862 · Cited by 6
“A promising solution is curriculum learning (CV) [14], [15], [16].”
Efficient evolutionary curriculum learning for scalable multi-agent reinforcement learning
Li, C., Liu, Y., Wang, J., Wang, Z., Wang, C., & Tian, Q. (2025) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-025-00215-y · Cited by 5
“Curriculum learning was initially widely applied in deep learning such as computer vision and natural language processing (Bengio et al. 2009).”
PPAAS: PVT and Pareto Aware Analog Sizing via Goal-conditioned Reinforcement Learning
Kim, S., Wang, Z., Lee, S., et al. (2025) · doi:10.1109/iccad66269.2025.11240795 · Cited by 4
“Another line of work focuses on curriculum learning, which aims to shape the goal distribution p ( g ) to guide the agent through a more structured learning progression [21, 22].”
Structure regularized self-paced learning for robust semi-supervised pattern classification
Gu, N., Fan, P., Fan, M., et al. (2018) · Neural Computing and Applications · doi:10.1007/s00521-018-3478-1 · Cited by 4
Self-paced Robust Deep Face Recognition with Label Noise
Zhu, P., Ma, W., & Hu, Q. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-16142-2_33 · Cited by 5
Caubrière, A., Tomashenko, N., Laurent, A., Morin, E., Camelin, N., & Estève, Y. (2019) · Interspeech (USB) · doi:10.21437/interspeech.2019-1832 · Cited by 54
“Basic idea of curriculum learning is to ”start small, learn easier aspects of the task or easier sub tasks, and then gradually increase the difficulty level” [10].”
BackgroundTakase, T. (2019) · Expert Systems with Applications · doi:10.1016/j.eswa.2019.06.017 · Cited by 4
Making Deep Neural Networks Robust to Label Noise: A Reweighting Loss and Data Filtration
Zhang, Z., Li, Y., Li, Y., & Qin, Y. H. (2019) · IEEE International Conference on Signal and Image Processing (ICSIP) · doi:10.1109/siprocess.2019.8868645 · Cited by 4
“Different from [5], Self-paced learning (SPL) [6] generates a binary reweighting factor, i.e. a factor representing whether an example is easy or hard, through the values of losses.”
BackgroundCooperative Behavior by Multi-agent Reinforcement Learning with Abstractive Communication
Tanda, J., Moustafa, A., & Itō, T. (2019) · IEEE International Conference on Agents (ICA) · doi:10.1109/agents.2019.8929151 · Cited by 4
“1) Curriculum Learning [14]: At the beginning of the training, agents choose an action randomly because there is no experience yet.”
BackgroundInfluentialDeduplication of Advertisement Assets Using Deep Learning Ensembles
Chakiat, A., Oli, N., & Modi, V. K. (2020) · International Conference on Machine Learning and Applications · doi:10.1109/icmla51294.2020.00145 · Cited by 5
“For the data collection, authors of SimNet use a novel online pair mining strategy (OPMS) which is inspired from Curriculum Learning [5].”
MethodClassification of Mammography based on Semi-supervised Learning
Sun, L., Wen, J., Wang, J., et al. (2020) · IEEE International Conference on Progress in Informatics and Computing · doi:10.1109/pic50277.2020.9350835 · Cited by 4
“Curriculum learning (CL) [28] achieves the goal of learning all knowledge of the model by simulating the mechanism of human knowledge learning from easy to hard.”
BackgroundSelf-paced hierarchical metric learning (SPHML)
Al-taezi, M. A., Zhu, P., Hu, Q., Wang, Y., & Al-badwi, A. (2021) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-021-01336-2 · Cited by 4
“(2) Supervised deep learning methods such as DNLML-ISA [2] and DDML [13], SPHML model outperforms the other deep learning supervised method.”
BackgroundMethodDiscovering the Arrow of Time in Machine Learning
Kasmire, J., & Zhao, A. (2021) · Information · doi:10.3390/info12110439 · Cited by 5
Toddler-Guidance Learning: Impacts of Critical Period on Multimodal AI Agents
Park, J., Park, K., Oh, H., Lee, G., Lee, M., Lee, Y., & Zhang, B. (2021) · INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION · doi:10.1145/3462244.3479932 · Cited by 4
Assessing the impact of minor modifications on the interior structure of GRU: GRU1 and GRU2
Yiğit, G., & Amasyalı, M. F. (2021) · Concurrency and Computation Practice and Experience · doi:10.1002/cpe.6775 · Cited by 5
Curriculum Optimization for Low-Resource Speech Recognition
Kuznetsova, A., Kumar, A. K., Drexler, J., & Tyers, F. M. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746674 · Cited by 6
“Complexity measures for data-driven learning include sentence/utterance length [16], language model score, n-gram size [5, 10], word frequency ranking, and sentence norm [13].”
BackgroundMethodInfluentialAdaptor: Objective-Centric Adaptation Framework for Language Models
Štefánik, M., Novotný, V. S., Groverová, N., & Sojka, P. (2022) · Annual Meeting of the Association for Computational Linguistics: System Demonstrations (ACL Demo) · doi:10.18653/v1/2022.acl-demo.26 · Cited by 7
Construction Method of Air Combat Agent Based on Reinforcement Learning
Mao, Y., Li, Q., He, J., Xia, Z., & Fei, A. (2022) · Lecture notes in electrical engineering · doi:10.1007/978-981-19-6052-9_10 · Cited by 4
Enhancing Semi-Supervised Learning with Cross-Modal Knowledge
Zhu, H., Lu, Y., Wang, H., Zhou, X., Ma, Q., Liu, Y., Jiang, N., Wei, X., Zeng, L., & Zhao, X. (2022) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3503161.3548026 · Cited by 5
“The essence of curriculum learning is to make the model learn easy samples before learning hard samples [2].”
BackgroundCDZoom: a human-like sequential zoom agent for efficient change detection in large scenes
Lin, Y., Wu, F., & Zhao, J. (2022) · Neural Computing and Applications · doi:10.1007/s00521-022-08096-2 · Cited by 3
Deleruyelle, A., Versari, C., & Klein, J. (2022) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2022.106454 · Cited by 4
“This learning phase is more challenging than the previous ones and it has proved much beneficial to execute it as part of a curriculum learning (CL) procedure [3].”
BackgroundWas that so Hard? Estimating Human Classification Difficulty
Hannemose, M. R., Sundgaard, J. V., Ternov, N. K., Paulsen, R. R., & Christensen, A. N. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-17721-7_10 · Cited by 5
DistilNAS: Neural Architecture Search With Distilled Data
Prabhakar, S. N., Deshwal, A., Mishra, R., & Kim, H.-S. (2022) · IEEE Access · doi:10.1109/access.2022.3224788 · Cited by 4
“In this work, we leverage concepts from curriculum learning [35] in order to accelerate the search stage of NAS.”
MethodTowards Disturbance-Free Visual Mobile Manipulation
Ni, T., Ehsani, K., Weihs, L., & Salvador, J. (2023) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00519 · Cited by 11
“Our two-stage training methodology reframes the single task learning (success without disturbance) into “curriculum learning” [7, 67] on a task sequence, where the first task (stage) is to succeed with disturbance, and the second task (stage) is to succeed without disturbance.”
MethodUCL: Unsupervised Curriculum Learning for Utility Pole Detection from Aerial Imagery
Abid, N., Kovács, G., Wedin, J., Paszkowsky, N. A., Faisal Shafait, B., & Liwicki, M. (2022) · International Conference on Digital Image Computing: Techniques and Applications (DICTA) · doi:10.1109/dicta56598.2022.10034610 · Cited by 5
“To tackle these limitations, the concept of Curriculum Learning (CL) (proposed by Bengio et al. [26]) has been used by some researchers in the machine learning domain [27].”
BackgroundUnderstanding Difficulty-Based Sample Weighting with a Universal Difficulty Measure
Zhou, X., Wu, O., Zhu, W., & Liang, Z. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26409-2_5 · Cited by 11
FANS: Fast Non-Autoregressive Sequence Generation for Item List Continuation
Liu, Q., Zhu, J., Wu, J., Wu, T., Dong, Z., & Wu, X.-M. (2023) · International World Wide Web Conference (WWW) · doi:10.1145/3543507.3583430 · Cited by 4
“3) To improve the inference quality of non-autoregressive generation, we propose to train our model with curriculum learning [1] by using easy samples and hard samples in a progressive manner.”
MethodGradual Language Model Adaptation Using Fine-Grained Typology
Fekete, M. R., & Bjerva, J. (2023) · Workshop on Research in Computational Linguistic Typology and Multilingual NLP (SIGTYP) · doi:10.18653/v1/2023.sigtyp-1.19 · Cited by 3
“Gradual model adaptation is also informed by principles of curriculum learning, which aims to find an ideal ordering of training instances in order to enhance LM learning (Bengio et al., 2009).”
MethodPCC: Paraphrasing with Bottom-k Sampling and Cyclic Learning for Curriculum Data Augmentation
Lu, H., & Lam, W. P. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.5 · Cited by 16
CPLT: Curriculum Pseudo Label Transformer for Domain Adaptive Object Detection in Foggy Weather
Zhang, G., Wang, L., Zhang, Z., & Chen, Z. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3325266 · Cited by 4
“Based on this analysis, we propose DAC module, a curriculum learning [62] strategy that considers both the ground truth class distribution and the pseudo label distribution by dynamically adjusting class-specific thresholds during the training process.”
MethodEntropy-based Optimization on Individual and Global Predictions for Semi-Supervised Learning
Zhao, Z. W., Zhao, M., Liu, Y., Yin, D., & Zhou, L. (2023) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3581783.3612567 · Cited by 7
“Alternatively, other studies [11, 46, 49] integrated methods such as curriculum learning [6], uncertainty evaluations [39], Sinkhorn-Knopp algo-rithm [17], to further polish the pseudo-labels.”
BackgroundIntrinsically Motivated High-Level Planning for Agent Exploration
Sartor, G., Oddi, A., Rasconi, R., & Santucci, V. G. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-47546-7_9 · Cited by 6
Downlink Beamforming Optimization via Deep Learning
Johnston, J., Liu, X.-Y., Wu, S., et al. (2023) · Allerton Conference on Communication, Control, and Computing · doi:10.1109/allerton58177.2023.10313512 · Cited by 4
“The term was coined in [7] where “the basic idea is to start small, learn easier aspects of the task or easier sub-tasks, and then gradually increase the difficulty level.””
BackgroundRethinking AI for Science: An Evolution From Data‐Driven to Data‐Centric Framework
Maskey, M. (2023) · Perspectives of Earth and Space Scientists · doi:10.1029/2023cn000222 · Cited by 11
“Curriculum learning (Bengio, et al., 2009) has also shown to improve performance.”
BackgroundDecentralized Multi-Agent Reinforcement Learning with Global State Prediction
Bloom, J., Paliwal, P., Mukherjee, A., & Pinciroli, C. (2023) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros55552.2023.10341563 · Cited by 11
“When training with a gate obstacle, we employ curriculum learning [41], with the opening distance of the gate starting equal to the vertical width of the environment (i.e, no obstacle).”
MethodJeong, Y., Yang, J., Choi, I. H., & Lee, J. (2024) · IEEE Access · doi:10.1109/access.2024.3373470 · Cited by 7
“In natural language projects, if an LLM suffers from narrow domain dataset issues - a lack of diversity, there are known solutions to mitigate this problem: fine-tuning [5], data augmentation [10], domain adaptation [6], curriculum learning [7] and active learning [8].”
BackgroundZhao, J., Xu, J., Xu, Y., et al. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3376539 · Cited by 4
“2) CurriculumofUser-LevelSampling: Curriculumlearning takes cues from human learning patterns, where individuals start with simpler samples and gradually tackle more complex ones [37].”
BackgroundPer-class curriculum for Unsupervised Domain Adaptation in semantic segmentation
Alcover‐Couso, R., SanMiguel, J. C., Escudero-Viñolo, M., & Carballeira, P. (2024) · The Visual Computer · doi:10.1007/s00371-024-03373-8 · Cited by 12
“To that end, we propose a curriculum-based learning (CL) strategy [16].”
InfluentialEvaluating the Effectiveness of Deep Learning Models for Foundational Program Analysis Tasks
Chen, Q., Yu, C., Liu, R., Zhang, C., Wang, Y., Wang, K., Su, T., & Wang, L. (2024) · Proceedings of the ACM on Programming Languages · doi:10.1145/3649829 · Cited by 7
“We note that our training workflow, especially the last two phases, resembles an influential learning strategy called curriculum training [Bengio et al. 2009], under which models are trained on data shuffled in the ascending order of difficulty levels.”
MethodTargeted Transfer Learning: Leveraging Optimal Transport for Enhanced Knowledge Transfer
Ahamed, S. F., Islam, K. A., & Shetty, S. S. (2024) · 2012 International Conference on Computing, Networking and Communications (ICNC) · doi:10.1109/icnc59896.2024.10556294 · Cited by 4
“There are several approaches available in transfer learning to resolve the difference between two domains, such as domain adaptation [1], data augmentation [4], curriculum learning [5], data selection, and data shift [6].”
BackgroundFeldmeier, P., & Fraser, G. (2024) · doi:10.1145/3663532.3664467 · Cited by 6
“Target statements are selected by querying the (CDG) [9] for statements that are direct children of already covered program statements, allowing Neatest to apply a variation of Curriculum Learning [5] where the game is explored iteratively by seeding initial generations with prior solutions.”
BackgroundHiER: Highlight Experience Replay for Boosting Off-Policy Reinforcement Learning Agents
Horváth, D., Martín, J. B., Erdős, G., Istenes, Z., & Moutarde, F. (2024) · IEEE Access · doi:10.1109/access.2024.3427012 · Cited by 5
“Following [34], in this work, we consider both the data collection and the data exploitation meth-ods as curriculum learning (CL) methods [34], [35], [36].”
BackgroundMethodResultInfluentialCurriculum meta-learning for zero-shot cross-lingual transfer
Doan, T., & Le, B. H. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.112238 · Cited by 4
Learning to Walk with Adaptive Feet
Scaldaferri, A., Angelini, F., & Garabini, M. (2024) · Robotics · doi:10.3390/robotics13080113 · Cited by 4
“As in [14], the rough terrain policies were trained in a game-inspired automatic curriculum fashion [37], starting from simpler terrain samples and increasing in complexity once the control policies allow the robots to traverse them [38].”
SpanGNN: Towards Memory-Efficient Graph Neural Networks via Spanning Subgraph Training
Gu, X., Li, H., Gao, S., et al. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70352-2_15 · Cited by 4
Learning to Compare Hardware Designs for High-Level Synthesis
Bai, Y., Sohrabizadeh, A., Ding, Z., Liang, R., Li, W., Wang, D., Ren, H. M., Sun, Y., & Cong, J. (2024) · ACM/IEEE Workshop on Machine Learning for CAD (MLCAD) · doi:10.1145/3670474.3685940 · Cited by 4
Monocular Depth Estimation on Adverse Weathers With Curriculum Domain Distribution Alignment
Zhang, J., Li, L., Yan, C. C., Ke, W., & Gong, Y. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3456097 · Cited by 8
DigiWeather: Synthetic Rain, Snow and Fog Dataset Augmentation
Nikolov, I. A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-71707-9_2 · Cited by 10
Unsupervised Feature Learning with Emergent Data-Driven Prototypicality
Guo, Y., Zhang, Y., Chen, Y., & Yu, S. X. (2024) · 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/cvpr52733.2024.02189 · Cited by 5
“Representative examples can be viewed as prototypes and used for interpretable machine learning [3], curriculum learning [2], and learning better decision boundaries [5].”
BackgroundData Distribution-Based Curriculum Learning
Chaudhry, S., & Sharma, A. (2024) · IEEE Access · doi:10.1109/access.2024.3465793 · Cited by 7
“Curriculum learning enhances training of a model by starting with simple concepts and then gradually introducing difficult concepts as training progresses [21], [22].”
BackgroundMethodChoi, T., Would, O., Salazar-Gomez, A., Liu, X., & Cielniak, G. (2024) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2024.109416 · Cited by 6
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.