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 3,401–3,500 of 5,240
Curriculumformer: Taming Curriculum Pre-Training for Enhanced 3-D Point Cloud Understanding
Fei, B., Luo, T., Yang, W., Liu, L., Zhang, R., & He, Y. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3406587 · Cited by 5
“Curriculum learning [17] is a learning paradigm inspired by human learning that involves starting from simple tasks and gradually increasing to more difficult ones.”
SICode: Embedding-Based Subgraph Isomorphism Identification for Bug Detection
Gong, Y., Nie, J., You, W., Shi, W., Huang, J., Liang, B., & Zhang, J. (2024) · IEEE International Conference on Program Comprehension (ICPC) · doi:10.1145/3643916.3646556 · Cited by 3
“…is strictly bounded to make sure of the model’s performance, i.e., the node number of 𝑃 0 / 𝑆 is less than ten and the node number of the target graphs 𝑃 𝑖 s/ 𝑁 s increases from 10 to 120 according to curriculum training scheme [4], and the same restriction is set to the testing samples.”
BackgroundDual-stage feedback network for lightweight color image compression artifact reduction
Chen, Z., He, X., Zhang, T., et al. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106555 · Cited by 2
DeepSarc-US: A Deep Learning Framework for Assessing Sarcopenia Using Ultrasound Images
Behboodi, B., Obrand, J., Afilalo, J., & Rivaz, H. J. (2024) · Applied Sciences · doi:10.3390/app14156726 · Cited by 3
Human-in-the-loop: infusing knowledge into neural networks
Suzuki, I., & Hartono, P. (2024) · IEEE International Conference on Mechatronics and Automation (ICMA) · doi:10.1109/icma61710.2024.10633061 · Cited by 3
“So far, the roles of humans in HITL include acting as a kind of oracle for AI in annotating difficult examples, arranging curriculums for the learning process of the AI [6], intervening during the learning process, making final decisions, and so on.”
BackgroundEnhancing Temporal Knowledge Graph Representation with Curriculum Learning
Liu, Y., Shen, Y., & Dai, Y. (2024) · Electronics · doi:10.3390/electronics13173397 · Cited by 4
SVR-AVT: Scale Variation Robust Active Visual Tracking
Zhang, B., Jin, S., Ouyang, Q., Yang, H., Zheng, Y., Fu, C., & Shi, D. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10650947 · Cited by 2
“Due to the difficulty of directly training opponents capable of generating multi-scale states, we adopt a phased curriculum learning manner [12], [25]–[27].”
BackgroundImproving Zero-Shot Coordination with Diversely Rewarded Partner Agents
Wu, P., Yang, Z., & Yang, P. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10649945 · Cited by 2
“The process is enhanced by adopting a simple curriculum learning (CL)[22] that begins with easier tasks and moves to harder ones, enabling agents to gradually adapt to diverse partners with varying rewards.”
BackgroundCross Domain Optimization for Speech Enhancement: Parallel or Cascade?
Wan, L., Liu, H., Shi, L., et al. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3468026 · Cited by 2
“Drawing on the curriculum learning concept [16], multi-stage mapping algorithms have emerged as a significant concept in speech front-end tasks.”
BackgroundImage-caption difficulty for efficient weakly-supervised object detection from in-the-wild data
Nebbia, G., & Kovashka, A. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00266 · Cited by 4
“With curriculum learning (CL) [2], we train models on “easy” concepts first, and then build up to more complex ones.”
$\gamma$-Razor: Hardness-Aware Dataset Pruning for Efficient Neural Network Training
Liu, L., Zhang, P., Liang, Y., Liu, J., Morra, L., Guo, B., Yu, Z., Zhang, Y., & Zeng, D. D. (2024) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2024.3453600 · Cited by 3
“Inspired by the learning principles of human beings, curriculum learning posits that DNNs benefit by starting from easy samples and then progressively increasing to more challenging ones [18].”
Optimal Transport of Diverse Unsupervised Tasks for Robust Learning from Noisy Few-Shot Data
Que, X., & Yu, Q. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72933-1_17 · Cited by 2
Neighbor Does Matter: Curriculum Global Positive-Negative Sampling for Vision-Language Pre-training
Huang, B., He, F., Wang, Q., Chen, H., Li, G., Feng, Z., Wang, X. Y., & Zhu, W. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3681502 · Cited by 6
“By incorporating curriculum learning[4, 39, 41, 47, 48], we progressively introduces more complex and challenging samples throughout the training process[6, 7, 21, 22, 32, 45, 49, 50].”
BackgroundDefying catastrophic forgetting via influence function
Gao, R., & Liu, W. (2024) · Artificial Intelligence · doi:10.1016/j.artint.2024.104261 · Cited by 4
Impact of Strategic Sampling and Supervision Policies on Semi-Supervised Learning
Roy, S., & Etemad, A. (2024) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2024.3502453 · Cited by 4
“Curriculum learning was popularized by [35] in the context of machine learning.”
BackgroundGaitW: Enhancing Gait Recognition in the Wild Using Dynamic Information
Thapar, D., Chaudhari, J., Manchanda, S., Nigam, A., & Arora, C. (2024) · Lecture notes in computer science · doi:10.1007/978-981-96-0885-0_2 · Cited by 4
“Curriculum Learning In recent years, curriculum learning [3,11] has received growing research interests in natural language processing [23,24,29,32,49,52,53] and computer vision [26, 44, 47].”
Sturm, F., Trat, M., Sathiyababu, R., Allipilli, H., Menz, B., Hergenroether, E., & Siegel, M. (2025) · Machine Vision and Applications · doi:10.1007/s00138-024-01638-9 · Cited by 2
“In the context of this work, however, a curricu - lum learning approach [1 ] is suggested for each batch.”
Deep learning of optimal exercise boundaries for American options
Kim, H., & Huh, J. (2024) · International Journal of Computer Mathematics · doi:10.1080/00207160.2024.2442585 · Cited by 2
“To enhance training stability and performance, we utilize curriculum learning [9,45], gradually expanding the learning domain.”
Accelerating deep learning with fixed time budget
Khan, M. A., Hamila, R., & Menouar, H. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10637-w · Cited by 4
“Third, using efficient training techniques such as curriculum learning [10, 11] without pruning the model parameters or eliminating training data samples, model training can be converged faster [5].”
Gradual Domain Adaptation via Normalizing Flows
Sagawa, S., & Hino, H. (2025) · Neural Computation · doi:10.1162/neco_a_01734 · Cited by 16
“They propose a method, following a similar concept to curriculum learning (Bengio et al., 2009), that starts with simple tasks and gradually introducing more challenging problems for learning.”
ArithmeticGPT: empowering small-size large language models with advanced arithmetic skills
Liu, Z., Zheng, Y., Yin, Z., Chen, J., Liu, T., Mi, T., & Luo, W. (2025) · Machine Learning · doi:10.1007/s10994-024-06681-1 · Cited by 5
“Inspired by Bengio et al.’s curriculum learning theory (Bengio et al., 2009; Wang et al., 2021), our supervised fine-tuning strategy tries to mimic the human math learning process that first learns to recognize mathematical symbols, notations and numerical expressions and then learns to understand…”
Curricular Subgoals for Inverse Reinforcement Learning
Liu, S., Qing, Y., Xu, S., Wu, H., Zhang, J., Cong, J., Chen, T., Liu, Y.-F., & Song, M. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3532519 · Cited by 6
“Núñez-Molina et al. [36] model the goal selection as an RL problem to utilize the Deep Q-Network framework [33] to select goals.”
Multi-frequency progressive refinement for learned inverse scattering
Melia, O., Tsang, O., Charisopoulos, V., Khoo, Y., Hoskins, J., & Willett, R. M. (2025) · Journal of Computational Physics · doi:10.1016/j.jcp.2025.113809 · Cited by 4
“Such a sequence can be constructed explicitly ( e.g. , by varying regularization levels) or implicitly; for example, curriculum learning [Bengio et al., 2009] progressively adjusts the training data distribution from “easy” to “hard” samples and has been used to train physics-informed neural…”
BackgroundA Plug-In Curriculum Scheduler for Improved Deformable Medical Image Registration
Shin, J., Kim, Y., & Lee, J. (2025) · IEEE Access · doi:10.1109/access.2025.3539890 · Cited by 4
“Bengio et al. (2009) further extends the idea to various vision and language tasks - where multi-stage curriculum strategies give rise to improved generalization and faster convergence.”
Automatic Parsing and Extraction of Structured Data from Fax Images
Rizan, R. U. B. (2024) · International Conference on Engineering and Emerging Technologies (ICEET) · doi:10.1109/iceet65156.2024.10913905 · Cited by 6
“We use a stage-wise training strategy inspired by [21].”
Basaran, B. D., Zhang, X., Matthews, P. M., & Bai, W. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-84525-3_5 · Cited by 4
Wang, Y., Li, H., Gong, M., Wu, Y., Gong, P., Qin, A. K., Xing, L., & Zhang, M. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3560329 · Cited by 5
“Bengio et al. [35] introduced CL, a learning paradigm inspired by human learning processes, where training starts with easy samples and progressively incorporates harder ones.”
Delving Into Invisible Semantics for Generalized One-Shot Neural Human Rendering
Lin, Y., Xu, X., Zhang, H., et al. (2025) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2025.3563229 · Cited by 2
“This scheme shares a similar spirit to curricular learning [50], [51], [52], which allows the model to start with learning easy tasks and gradually adapt to more difficult tasks for better optimization.”
Abbas, A., Gaber, M. M., & Abdelsamea, M. M. (2025) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2025.3562620 · Cited by 2
“Just as we acquire skills and knowledge in a sequential manner, starting with the basics and gradually progressing to more complex topics, this approach to machine learning follows a similar pattern [9].”
Awais, C. M., Reggiannini, M., & Moroni, D. (2025) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw65960.2025.00171 · Cited by 4
“Curriculum learning has proven beneficial [1,28] in various domains of deep learning, including natural language processing [4,33,43], speech applications [14,29,39], computer vision tasks such as object detection [19, 27], image segmentation [5, 31], and image classification [6, 21, 30].”
Petchpol, K., & Boongasame, L. (2025) · Applied Computational Intelligence and Soft Computing · doi:10.1155/acis/3874667 · Cited by 6
Towards Polyp Counting in Full-Procedure Colonoscopy Videos
Parolari, L., Cherubini, A., Ballan, L., & Biffi, C. (2025) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi60581.2025.10981023 · Cited by 6
“To address this, inspired by curriculum learning [15], we gradually incorporate such pairs into training by increasing their sampling probability in later training stages.”
CyclicAugment: Optimized Medical Image Analysis via Adaptive Augmentation Intensity
Kim, M.-J., Chae, J.-W., & Cho, H. (2025) · IEEE Access · doi:10.1109/access.2025.3570899 · Cited by 4
“As training progresses, increasingly complex data are incorporated to enhance learning stability and generalization performance [26], [27], [28].”
Sun, M., & Qin, S. (2025) · Applied Sciences · doi:10.3390/app15105686 · Cited by 2
Ren, L., Xian, Y., Li, S., et al. (2025) · Astrodynamics · doi:10.1007/s42064-023-0184-2 · Cited by 2
“Based on the idea of curriculum learning (CL) [49, 50], a multi-level NN-learning strategy based on CL is proposed.”
Ting, J. Y. C., & Barnard, A. S. (2025) · Advanced Intelligent Discovery · doi:10.1002/aidi.202500033 · Cited by 5
Revisiting Source-Free Domain Adaptation Object Detection in Thresholds
Dong, Y., Li, C., Xie, Y., & Li, Z. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3590905 · Cited by 4
“To improve the quality of pseudo-labels, we draw inspiration from the curriculum learning [28], [29] and few-shot learning [30], [31].”
Knowledge-Enhanced Prompt-Tuning With Contrastive Learning for Emotion Recognition in Conversations
Gao, Q., Cao, J., Cao, B., Guan, X., & Liu, B. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3592325 · Cited by 2
“SPCL-CL-ERC [50] perform context encoding based on Sim-CSE [51] with a supervised prototypical contrastive learning loss and use a curriculum learning [60] strategy to reorganize the datasets.”
LLMs for Microservice Generation: Capabilities, Challenges, and Advancements
Spista, R., Crispo, B., Giorgini, P., Marchetto, A., & Riccardi, G. (2025) · IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT) · doi:10.1109/iaict65714.2025.11100313 · Cited by 5
Meng, L., Li, Y., Wei, W., & Yang, C. (2025) · Symmetry · doi:10.3390/sym17091386 · Cited by 4
Structure-Preserving Histopathological Stain Normalization via Attention-Guided Residual Learning
Madusanka, N., Padmanabha, P., Guruge, K., & Lee, B.-I. (2025) · Bioengineering · doi:10.3390/bioengineering12090950 · Cited by 2
“The curriculum learning strategy provides substantial improvements in edge preservation loss and moderate improvements in structural similarity, while attention gates contribute moderately to both metrics, confirming the importance of all components for optimal training dynamics [49].”
FeUDA-BERT: Federated URL Domain-Aware BERT for Phishing URL Detection
Rashid, F., Doyle, B., & Seneviratne, S. (2025) · Conference on Local Computer Networks · doi:10.1109/lcn65610.2025.11146376 · Cited by 4
HiFAR: Multi-Stage Curriculum Learning for High-Dynamics Humanoid Fall Recovery
Chen, P., Wang, Y., Luo, C., Cai, W., & Zhao, M. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11245953 · Cited by 7
“Multi-Stage Curriculum Learning (MSCL) represents a training methodology that incrementally escalates the complexity of learning tasks by introducing supplementary constraints or objectives at each phase [20].”
Toward Exploring End-to-End Learning Algorithms for Autonomous Aerial Machines
Krishnan, S., Boroujerdian, B., Faust, A., et al. (2019) · Cited by 2
Astudillo, P., Mortier, P., Beule, M. D., & wyffels, F. (2020) · International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC) · doi:10.5220/0008948900370045 · Cited by 3
“Curriculum learning (Bengio et al., 2009) is inspired by the human education system, where students start with a straightforward concept before learning more advanced concepts.”
BackgroundUnified Gradient Reweighting for Model Biasing with Applications to Source Separation
Tzinis, E., Bralios, D., & Smaragdis, P. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414071 · Cited by 2
“Curriculum learning starts by using higher weights on “easier” examples while gradually shifting to a uniform weight distribution and has been shown to significantly reduce training time as well as increasing generalization capabilities of models [16].”
BackgroundMethodInfluentialUnihanLM: Coarse-to-Fine Chinese-Japanese Language Model Pretraining with the Unihan Database
Xu, C., Ge, T., Li, C., & Wei, F. (2020) · Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (AACL-IJCNLP) · doi:10.18653/v1/2020.aacl-main.24 · Cited by 5
“Learning (Bengio et al., 2009), which smooths the training and help the model generalize better.”
BackgroundMethodZhang, Y., Shi, D., Wu, Y., Zhang, Y., Wang, L., & She, F. (2021) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-030-67537-0_36 · Cited by 2
Token-wise Curriculum Learning for Neural Machine Translation
Chen, L., Jiang, H., Liu, X., He, P., Chen, W., Gao, J., & Zhao, T. (2021) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2021.findings-emnlp.310 · Cited by 4
“Aside from previous approaches, curriculum learning (Bengio et al., 2009) is proposed to address the data insufficiency issue via utilizing the limited data more efficiently (Zhang et al., 2019b).”
MethodLifelong Classification in Open World With Limited Storage Requirements
Bi, W., Yang, C., Li, X., & Chen, J. (2021) · Neural Computation · doi:10.1162/neco_a_01391 · Cited by 2
“The reason for the use of L 1 -norm is that it can lead to the sparse weights (Bengio et al., 2009) and reduces the cost of the storage of weights.”
BackgroundBurdick, L., Kummerfeld, J. K., & Mihalcea, R. F. (2021) · Mathematics · doi:10.3390/math9182234 · Cited by 3
Uncertainty-Aware Self-paced Learning for Grammatical Error Correction
Dang, K., Xie, J., Liu, J., & Chen, S. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-88480-2_59 · Cited by 2
Restoring Critical Loads In Resilient Distribution Systems using A Curriculum Learned Controller
Zhang, X., Eseye, A. T., Reynolds, M. J., Knueven, B., & Jones, W. B. (2021) · IEEE Power Engineering Society General Meeting · doi:10.1109/pesgm46819.2021.9637834 · Cited by 2
Development and Comparison of Scoring Functions in Curriculum Learning
Kesgin, H. T., & Amasyalı, M. F. (2022) · IEEE International Conference on Computing and Machine Intelligence (ICMI) · doi:10.1109/icmi55296.2022.9873743 · Cited by 3
“classification [4] and artificial data sets [5].”
MethodBehavioral Optimization in a Robotic Serial Reaching Task Using Predictive Information
Sen, D., de Kleijn, R., & Kachergis, G. (2022) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2022.3176459 · Cited by 2
“of convergence and the quality of local minima obtained in nonconvex optimization landscapes [34], [35].”
BackgroundToward Congestion Control in Lossy Networks via Local Queue-Management Policies
Forero, P. A., Zhang, P., & Radosevic, D. (2022) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom55135.2022.10017987 · Cited by 2
“This motivated us to refine the training method with ideas from curriculum learning which suggests to gradually increase the complexity of the environment during the training process [22].”
MethodParamCrop: Parametric Cubic Cropping for Video Contrastive Learning
Qing, Z., Huang, Z., Zhang, S., Tang, M., Gao, C., Jin, R., Ang, M. H., & Sang, N. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3244126 · Cited by 3
“Inspired by curriculum learning [22], we slightly increase the difficulty in the later stage of contrastive training by increasing the central distance between two cropped views (Fig.”
BackgroundAmbiLearn: Enhancing the Learning Environment for Primary School Education
Hyndman, J., Lunney, T., & Kevitt, P. M. (2012) · Lecture notes in computer science · doi:10.1007/978-3-642-33687-4_20 · Cited by 2
The Analysis of University Curriculum Setting Based on KNIME
Chen, G., & Yang, J. (2012) · IERI Procedia · doi:10.1016/j.ieri.2012.06.168 · Cited by 2
Smith, J. R. (2013) · IEEE Multimedia · doi:10.1109/mmul.2013.39 · Cited by 2
A Cognitive Model for Generalization during Sequential Learning
Gupta, A., Vig, L., & Noelle, D. C. (2011) · Journal of Robotics · doi:10.1155/2011/617613 · Cited by 2
“In robotics and machine learning, though, a number of attempts have been made in analysing enhanced learning through a sequence of training environments [30–35].”
BackgroundSingh, M., Greenberg, C., Oualil, Y., et al. (2016) · International Conference on Computational Linguistics · Cited by 2
META-LEARNING AND ALGORITHM SELECTION WORKSHOP AT ECAI 2014
Vanschoren, J., Brazdil, P., Soares, C., & Kotthoff, L. (2014) · Cited by 4
Classification of Ciliary Motion with 3D Convolutional Neural Networks
Lu, C., & Quinn, S. P. (2017) · doi:10.1145/3077286.3077303 · Cited by 2
“We attempted to mitigate this effect of misleading examples by following Bengio’s [1] principle of curriculum learning, opting to train only on patients labeled 1 and 4 (trading quantity of training examples for more canonical and distinct motion examples).”
MethodA Deep Learning Architecture for Histology Image Classification with Curriculum Learning
Kao, C.-Y., Madduri, M., & McMillan, L. (2017) · Lecture notes in computational vision and biomechanics · doi:10.1007/978-3-319-68195-5_122 · Cited by 2
“, 2014) shows that curriculum learning (Bengio et al., 2009b; Kumar et al., 2010b) can use an intermediate hint feature layer to steer difficult to learn objective functions towards effective solutions.”
BackgroundInfluentialModifications and Extensions to a Feed-Forward Neural Network
Skansi, S. (2018) · Undergraduate topics in computer science · doi:10.1007/978-3-319-73004-2_5 · Cited by 2
Liu, J., Du, A., Wang, C., Zheng, H., Wang, N., & Zheng, B. (2018) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr.2018.8546037 · Cited by 3
“Curriculum learning [32], [33] gains a great concern in the machine learning field recently, which is a new learning strategy that imitates the human learning process that learns from easy to complex samples.”
BackgroundAutonomous Curiosity for Real-Time Training Onboard Robotic Agents
Teng, E., & Iannucci, B. (2019) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2019.00163 · Cited by 4
“To stabilize training, we use curriculum learning [3], gradually increasing nobs from 10 to 20, where 20 produces scenes which are mostly obstructed.”
MethodPollack, T., & Kampen, E.-J. V. (2020) · AIAA SciTech Forum · doi:10.2514/6.2020-2100 · Cited by 2
“The proposed framework is based on the idea that for non-convex optimization tasks, the order in which training samples are presented can be used to avoid local minima [14, 28].”
BackgroundInfluentialMulti-Agent Deep Reinforcement Learning for Cooperative Driving in Crowded Traffic Scenarios
Park, J., Min, K., & Huh, K. (2019) · International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) · doi:10.1109/ispacs48206.2019.8986374 · Cited by 2
EEG-Based Driver Drowsiness Estimation Using Self-Paced Learning with Label Diversity
Xu, Y., & Wu, D. (2019) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci44817.2019.9002753 · Cited by 2
“Instead of transferring information from the auxiliary subjects, this paper adopts another strategy called curriculum learning [18], or selfpaced learning (SPL) [19], which helps avoid local minima and hence improves the generalization performance.”
BackgroundUnsupervised Learning in Space and Time over Several Generations of Teacher and Student Networks
Leordeanu, M. (2020) · Advances in computer vision and pattern recognition · doi:10.1007/978-3-030-42128-1_7 · Cited by 2
Label Noise Reduction Without Assumptions
Wei, J. (2020) · Cited by 3
Designing Curriculum for Deep Reinforcement Learning in StarCraft II
Hao, D., Sweetser, P., & Aitchison, M. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-64984-5_19 · Cited by 4
Coevolutionary Deep Reinforcement Learning
Cotton, D., Traish, J. M., & Chaczko, Z. (2020) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci47803.2020.9308290 · Cited by 4
“These pressures address self-play training instabilities by combining successful adaptations analogous to curriculum learning from supervised learning, in which it was found that the order training data was presented to neural networks influenced their convergence speed [11, 12].”
BackgroundImage inpainting using scene constraints
Hedjazi, M. A., & Genç, Y. (2021) · Signal Processing Image Communication · doi:10.1016/j.image.2021.116148 · Cited by 2
Samtani, P., Leiva, F., & Ruíz-del-Solar, J. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636046 · Cited by 3
“Curriculum-learning, or the division of a task in a curriculum [19], [20], has been used in multi-agent RL settings.”
BackgroundInstance Selection for Online Updating in Dynamic Recommender Environments
Thanthriwatta, T., & Rosenblum, D. S. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-75765-6_49 · Cited by 2
“In contrast to curriculum learning [3], where the model learning is regulated by gradually including easy to complex samples based on different types of heuristics, SPL incorporates parameters into model learning to support the selection process.”
BackgroundEffective Use of Target-side Context for Neural Machine Translation
Mino, H. (2021) · Journal of Natural Language Processing · doi:10.5715/jnlp.28.731 · Cited by 2
A patch-based and multi-instance learning strategy for pneumothorax classification on chest X-rays
Tian, Y., Chang, Y., & Yang, X. (2021) · Journal of Physics Conference Series · doi:10.1088/1742-6596/1976/1/012030 · Cited by 2
“The method is inspired by the curriculum learning [15] approach and multi-instance learning (MIL) [16] strategy.”
MethodSuzuki, S., Takeda, S., Tanida, R., Kimata, H., & Shouno, H. (2021) · Proceedings - International Conference on Image Processing · doi:10.1109/icip42928.2021.9506696 · Cited by 2
“In an extended study, Sinha et al. [8] determined the strength of blur filters in a curriculum manner [10] and demonstrated that blur filters work well for many image recognition tasks.”
BackgroundRLP-Net: A Recursive Light Propagation Network for 3-D Virtual Refocusing
Shin, C., Ryu, H., Cho, E., et al. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-87231-1_18 · Cited by 2
Direct Oriented Ship Localization Regression in Remote Sensing Imagery with Curriculum Learning
Guo, W., Chen, H., Zhang, Z., et al. (2021) · IEEE International Geoscience and Remote Sensing Symposium (IGARSS) · doi:10.1109/igarss47720.2021.9554670 · Cited by 2
“In this paper, inspired recently proposed curriculum Learning framework[10], we design a curriculum learning multi-task training objective formatted as where L mask is the loss for the region mask prediction task, L loc for the localization parameter regression task, and α ( l ) is where ε is a…”
MethodTest Harder than You Train: Probing with Extrapolation Splits
Kunz, J., & Kuhlmann, M. (2021) · BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP) · doi:10.18653/v1/2021.blackboxnlp-1.2 · Cited by 8
Curriculum Learning to Handle Extreme Class Imbalance for Acoustic Modeling of Forest Elephant Calls
Gomes-Selman, J. M., Demir, N., Wrege, P. H., et al. (2021) · IEEE International Conference on Machine Learning and Applications (ICMLA) · doi:10.1109/icmla52953.2021.00168 · Cited by 2
Improved training of deep convolutional networks via minimum-variance regularized adaptive sampling
Rojas‐Domínguez, A., Valdez, S. I., Ornelas-Rodríguez, M., & Carpio, M. (2022) · Soft Computing · doi:10.1007/s00500-022-07131-7 · Cited by 3
“In yet another approach inspired by how humans and animals learn, Bengio et al. (2009) presented training instances to learning models in gradually increasing amounts and degree of complexity, achieving faster convergence and better generalization in a variety of toy problems, recognition of…”
BackgroundSelf-paced latent embedding space learning for multi-view clustering
Li, H., Ren, Z., Zhao, C., et al. (2022) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-022-01600-z · Cited by 2
“Enlightened by curriculum learning, the work in [32] proposes self-paced learning (SPL).”
BackgroundTzur, M., Paryanti, G., & Sadot, D. (2022) · IEEE Transactions on Communications · doi:10.1109/tcomm.2022.3179765 · Cited by 2
Option Discovery for Autonomous Generation of Symbolic Knowledge
Sartor, G., Zollo, D., Mayer, M. C., Oddi, A., Rasconi, R., & Santucci, V. G. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-08421-8_11 · Cited by 2
Soft Actor-Critic with Inhibitory Networks for Retraining UAV Controllers Faster
Choi, M., Filter, M., Alcedo, K., Walker, T. T., Rosenbluth, D., & Ide, J. S. (2022) · International Conference on Unmanned Aircraft Systems (ICUAS) · doi:10.1109/icuas54217.2022.9836052 · Cited by 3
Zhou, Q., Wang, R., Zeng, G., Fan, H., & Zheng, G. (2022) · Medical Image Analysis · doi:10.1016/j.media.2022.102607 · Cited by 2
On the Verge of Solving Rocket League using Deep Reinforcement Learning and Sim-to-sim Transfer
Pleines, M., Ramthun, K., Wegener, Y., Meyer, H., Pallasch, M., Prior, S., Drogemuller, J., Büttinghaus, L., Rothemeyer, T., Kaschwig, A., et al. (2022) · IEEE Conference on Games (CoG) · doi:10.1109/cog51982.2022.9893628 · Cited by 6
“…leading to an even more delayed reward signal and more challenging credit assignment • More states have to be explored by the agent • Even more accurate touches on the ball have to be made to score To overcome these problems, curriculum learning [32] and reward shaping [33] can be considered.”
BackgroundGathering Physical Particles with a Global Magnetic Field Using Reinforcement Learning
Konitzny, M., Lu, Y., Leclerc, J., Fekete, S. P., & Becker, A. T. (2022) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9982256 · Cited by 6
“To slightly reduce the impact of altered observations on the overall training time, we decided to make use of curriculum learning [18] and introduce noise gradually during training: The amount of noise is linearly increased from zero after half of the training is finished and reaches its maximum value at 90% of the tr…”
MethodLearning to Box: Reinforcement Learning using Heuristic Three-step Curriculum Learning
Rho, H., Yu, Y., & Lee, K. (2022) · International Conference on Control, Automation and Systems (ICCAS) · doi:10.23919/iccas55662.2022.10003778 · Cited by 3
“Bengio et al. [13] introduced curriculum learning (CL) as a machine learning concept to improve the performance of supervised learning.”
BackgroundSemi-Supervised Learning for Medical Image Classification Based on Anti-Curriculum Learning
Wu, H., Sun, J., & You, Q. (2023) · Mathematics · doi:10.3390/math11061306 · Cited by 6
“[39], the basic idea of which is to make the model learn from easy samples to hard samples.”
BackgroundImproving Imbalanced Text Classification with Dynamic Curriculum Learning
Zhang, X., Wang, J., Cheng, N., & Xiao, J. Z. (2022) · International Conference on Mobility, Sensing and Networking (MSN) · doi:10.1109/msn57253.2022.00167 · Cited by 4
“CL [6], which mimics the human learning process from easy to hard concepts, can improve the generalization ability and convergence rate of deep neural networks.”
BackgroundA Multi-Stage Triple-Path Method For Speech Separation in Noisy and Reverberant Environments
Mu, Z., Yang, X., Yang, X., et al. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10096086 · Cited by 2
“To address these issues, inspired by curriculum learning concept [12, 13], we design a single-model multi-stage training method, which can not only train an end-to-end network to simplify the training process but also outperforms single-stage methods with the same model size.”
Teaching Deep Learners to Generalize
Aggarwal, C. C. (2023) · Springer eBooks · doi:10.1007/978-3-031-29642-0_5 · Cited by 2
Ghosh, S., Menon, R. R., & Srivastava, S. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.467 · Cited by 8
Learning modular language-conditioned robot policies through attention
Zhou, Y., Sonawani, S. D., Phielipp, M., Amor, H. B., & Stepputtis, S. (2023) · Autonomous Robots · doi:10.1007/s10514-023-10129-1 · Cited by 16
“In this section, we describe hierarchical modularity—an algorithm for training hierarchies of modules which is inspired by curriculum learning (Bengio et al., 2009).”
Method
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.